text
stringlengths 4
2.52M
| meta
dict |
|---|---|
RS-1: Pediatric dosimetry of
18
F-FDG whole body PET/CT scans
**Aruna Kaushik, Anupam Mondal^1^, Rajnish Sharma^1^, Maria D'Souza^1^, Abhinav Jaimini^1^, Dinesh Singh^1^, Anjani Tiwari, AK Mishra, BS Dwarakanath^2^**
Division of Cyclotron and Radiopharmaceutical Sciences, ^1^Division of PET Imaging, ^2^Division of Radiation Biosciences, Institute of Nuclear Medicine and Allied Sciences, Brig. S.K. Mazumdar Marg, Timarpur, Delhi, India
**Introduction:** A combined ^18^F-FDG (^18^F-2-deoxy-D-glucose) positron emission tomography/ computed tomography (PET/CT) scan provides both the metabolic information from FDG-PET and anatomic information from CT in a single examination. The use of PET/CT for management of malignancies in children has increased over the past few years. This raises an important consideration of radiation exposure in children since they are relatively more radiosensitive than adults and also have a potential for a longer life thereby increasing the probability of manifestation of late radiation effects; particularly cancer. Unfortunately, the data regarding the doses received by children from undergoing such examinations is scarce. The present study aims at estimating the effective doses to paediatric patients from whole body ^18^F-FDG PET/CT studies.
**Objective:** The purpose of the study is to estimate the radiation doses to children from undergoing whole body PET/CT scans using ^18^F-FDG.
**Materials and Methods, Subjects:** A random sample of hundred (age range: 0-2, 2-6, 6-12 and 12-18 years) paediatric patients from ^18^F-FDG PET/CT studies was included for dosimetry estimates. The average weight (in kg) of the children for the above mentioned age range was 10.3±2.1, 16.3±4.1, 29.6±8.8, 42.2±10.7 and height (in cms) was 87±8.9, 104.2±10.5, 130.6±14.6, 155.5±10. The average amount of activity (in MBq) of ^18^F-FDG injected for the different age ranges was 150±36, 199.15±45.8, 246.7±49.7, and 288.2±49.1 respectively. The use of patient\'s data for the purpose of estimating radiation doses was approved by the institutional human ethics committee.
**Protocol for whole body PET/CT scan:** The protocol for whole body ^18^F-FDG PET/CT examinations comprised of (a) a topogram, or scout scan, for positioning; (b) a spiral CT scan for attenuation correction; (c) a PET scan over the same axial range as the CT scan. The patients were scanned from the top of the skull to mid thigh.
**Internal dosimetry:** The effective doses are computed by using the dose coefficients and tissue weighting factors recommended by ICRP. Effective doses, E were estimated by:
E = A • Γ~E~
Where
A is the administered activity and Γ~E~ is the dose coefficient for the effective dose.
**External dosimetry (CT):** The effective dose from the CT component was computed using CT-Expo software on the basis of the scan parameters and the characteristics of the model of CT system, Lightspeed 16 stored in the software data base. The scan parameters used in the spiral CT scan of whole-body ^18^F-FDG PET/CT studies were 120 kV, 110 mA, 0.8 s scan duration, beam width 10 mm, reconstructed slice thickness 3.75 mm, Table feed 17.5 mm and pitch 1.75:1.
**Results and Conclusion:** The total effective dose among the paediatric patients of the combined PET/CT studies, calculated by summing the effective doses of PET and CT scanning from ^18^F-FDG whole body PET/CT examination was 24.75±3.4 mSv, 20.05±2.6 mSv, 18.3±1.8 mSv and 16.3±1.2 mSv for the age ranges 0-2, 2-6, 6-12 and 12-18 years respectively. The dose received by the children in the age group 12-18 years is comparable to the adult doses reported by various studies. However, the dose received by the children in the age range 0-2 years is 34%, 2-6 years is 18% and 6-12 years is 10.9% more than the dose received by children in the age range 12-18 years. This may be due to higher metabolic rate, higher radiosensitivity and relatively more activity administered per unit mass of the body to the paediatric patients in the lower age groups. Higher effective dose to children in these age groups results in relatively higher risk of radiation as compared to adults. However, the effective dose and in turn the radiation risk to paediatric patients can be reduced by optimising the protocol for PET/CT scanning that involves administering the amount of PET pharmaceutical as per their body weight and also modulating the CT scan parameters according to the size and weight of the patient.
RS-2: Radiological safety of bone mineral densitometry equipment
**Santosh Kumar, Aruna Kaushik, Maria D'Souza, Dinesh Singh, Rajnish Sharma, Anupam Mondal**
Division of PET Imaging, Institute of Nuclear Medicine and Allied Sciences, Brig. S.K. Mazumdar Marg, Timarpur, Delhi, India
**Introduction:** Osteoporosis is a highly prevalent disease leading to increased risk of bone fractures. Bone mineral densitometry (BMD) is a well accepted clinical tool for the diagnosis and management of osteoporosis. There are several different modalities for BMD such as Dual Energy X-ray Absorptiometry, Quantitative Ultrasound, Radiographic Absortiometry and Quantitative Computerized Tomography. Measurement of bone mineral density (BMD) by dual energy X-ray absorptiometry (DXA) is now well established as the method of choice for osteoporosis assessment.
**Objective:** This study was conducted to assess the radiation dose to patients and staff from the standard scan modes.
**Materials and Methods:** We have conducted a survey to determine the whole body radiation dose received specifically through operation of the dual energy X-ray absorptiometry (DXA) system Lunar Prodigy bone densitometer, M/s GE Healhcare installed in our institute. This is operated by departmental technologist staff working in rotation. Each patient routinely receives a standard DXA scan of the whole body, spine, hip region and wrist. The scan parameters are 76 kV, 0.15 mA for the whole body, 76 kV, 3mA for spine, 76 kV, 3 mA for the hip region and 76 kV, 0.15 mA for the wrist. The radiation levels were measured at various room locations \[[Figure 1](#F1){ref-type="fig"}\]. The patient doses were observed from the values displayed on the Lunar Prodigy console for each scan.
{#F1}
**Results:** The radiation levels measured at various locations in the room housing the BMD system are tabulated in [Table 1](#T1){ref-type="table"}. The gamma radiation level measured at the locations of interest was less than the permissible limit of 10 μSv/h for working hours. The dose to patients from the four scan modes are tabulated in [Table 2](#T2){ref-type="table"}. The doses to the patients are much below the standard radiographic examinations.
**Conclusion:** The radiation levels at various locations in the room housing the equipment are much below the permissible levels. The equipment is safe from radiation safety standpoint. However, receiving unnecessary exposure by standing near the couch should be avoided and ALARA (As low as Reasonably Achievable) principle should be followed. The radiation dose received by patients and the risk of cancer from BMD is much lower as compared to other radiographic procedures.
######
Radiation level at various locations in BMD room
######
Radiation exposure with scanning modes
RS-3: Radiopharmacy contamination in nuclear medicine -- A survey report
**Anchal Ghai, Somnath Mardi, Pradeep Kumar, Sarika, BR Mittal, B Singh**
Department of Nuclear Medicine, PGIMER, Chandigarh, India
**Objectives:** To conduct the radiation survey of the category IV Nuclear Medicine department and to further measure and compare the level of loose contamination in areas with high levels of exposure at two different time points (morning and evening) using wipe test.
**Materials and Methods:** The radiation monitoring survey of thes Department of Nuclear Medicine was performed using G.M. counter based survey meter. The survey was performed in 12 rooms of the department. The room exposure levels were recorded in mR/hr at three different time intervals namely, before the start of day\'s work, during work and at the end of the day\'s work. Wipe testing was done in the areas showing high exposure levels twice a day, in morning and evening. Requirements for wipe testing were cotton swabs dipped in alcohol, forceps, a pair of gloves, test tubes with stand and well counter with calibrating sources. An area of 936.36 cm^2^ was swiped for each wipe sample collected. The data was collected for 30 different days.
**Results:** The average exposure in the non radioactive areas was found to be equal to background. The daily average exposure in the controlled areas of the department i.e hot lab and RIA lab were 0.174 mR/hr and 0.063 mR/hr respectively and that in the controlled supervised areas was almost equivalent to background. Wipe testing was performed at 7different locations in the hot lab (near fumehood, injection bed, L-bench 1, L-bench 2, sink, centre of lab and background). Injection bed, sink and area outside the door of radiopharmacy showed high contamination levels as compared to the recommended levels.
**Conclusion:** As a part of surveillance program of Nuclear Medicine laboratories, daily radioactivity contamination monitoring using wipe test in the areas of concern especially in extremely busy nuclear medicine departments (with high patient\'s load) will provide 'guidance' for careful compliance with radioactivity handling practices.
RS-4: Radiation exposure during leukocyte labelling process
**Sarika, A Bhattacharya, R Kochhar^1^, BR Mittal**
Department of Nuclear Medicine, ^1^Department of Gastroenterology, Postgraduate Institute of Medical Education and Research, Chandigarh, India
**Objective:** Leukocyte labelled imaging is considered a gold standard technique for the diagnosis of the infection. However, the radiolabeling of the white blood cells is a time consuming process, which requires open handling of the radio-nuclides for a considerable period of time leading to high radiation exposure to the personnel. In this study an attempt was made to record the total exposure received and the total time taken for the leukocyte labelling using two different radio-nuclides.
**Materials and Methods:** The average exposures on the workbench during the various steps of the radiolabeling were recorded using a survey meter. The personnel exposure was recorded using a pocket dosimeter.
**Results:** The average total time taken for the leukocyte labelling using 99m-Tc HMPAO and ^18^F-FDG were 23.08 and 35.42 minutes respectively. The mean radiation exposure received per radiolabeling procedure for 99m-Tc HMPAO and ^18^F-FDG were 2.04 microSievert and 6.44 microSievert respectively. The average exposures on the workbench during the various steps of the radiolabeling are given in the [Table 1](#T3){ref-type="table"}. The time taken for the ^18^F-FDG leukocyte labelling was higher than the Tc99m HMPAO leukocyte labelling due to longer incubation period required for the former process. The mean radiation exposure received is also higher in the ^18^F-FDG leukocyte labelling process due to the higher gamma ray dose constant of the ^18^F radionuclide.
######
Exposure (mr/h) working bench
**Conclusion:** The radiolabeling of leukocytes by any radionuclide can be safely performed using the basic radiation safety principle of TDS. The study demonstrated the safe leukocyte radiolabeling with two different radio-nuclides using the same set of radiation shielding equipments. However the radiation exposure is relatively higher in the ^18^F-FDG leukocyte labelling process than the 99m-Tc HMPAO leukocyte labelling process.
RS-5: Radiological safety in using N-13 radioisotope in the PET facility in India
**Ajay Kumar Gocher, Pankaj Tandon**
Radiological Safety Division, Atomic Energy Regulatory Board, Mumabi, India
**Introduction:** Positron emission tomography (PET) is the most viable diagnostic tool presently available to diagnose the functional abnormality of the organ in the human body. The most commonly used PET radio pharmaceutical is FDG labelled with F-18 radioisotope having reasonably good half life (110 min), making it available to the PET installation. These PET installations are approved by Atomic Energy Regulatory Board (AERB) for calculated work load (pt/week) based on the shielding adequacy of the facility. In PET installation mostly work load is limited by uptake room because patients have to spend more time in uptake room than imaging room. For some other study (Myocardial perfusion study) Nuclear Medicine physician require to administer N-13 labelled radiopharmaceutical to the patient having half life of 10 min. Because of its short half life and requirement of study, activity of 370-1110 MBq needs to be administered within the imaging room. In this case the patient need not go to the uptake room. Therefore, there is a need to re- calculate the work load (pt/week) for imaging room for N-13 administered patient in the approved PET installation using F-18.
**Objective:** To calculate the work load (pt/week) for N-13 administered patient in the PET-CT installation and verify, whether the workload for using N-13 is same as the workload for using F-18.
**Materials and Methods:** Typical administered activity of N-13 labelled compound is around 370-1110 MBq and maximum scan time is 20 min including patient set-up on the couch. The work load has been calculated as per AAPM TG-108 report.
**Result and Conclusion:** From the theoretical calculations it has been noted that the work load for using N-13 will not be same as using F-18. One may practice with N13 labelled compound in the installed PET-CT but keep in mind about the reduction in the work load by keeping the radiation level with in the permissible limit.
RS-6: High radiation area survey during the running of the Cyclotron
**Amandeep Kaur, Sarika Sharma, BR Mittal**
Department of Nuclear Medicine and PET, PGIMER, Chandigarh, India
**Introduction:** Routine cyclotron survey is an important part of overall radiation safety in the cyclotron facility. This procedure is routinely performed by the Medical Physicist/RSO during the operation of the cyclotron. This helps to ensure the proper radiation safety and prevents unjustified radiation exposure to the working staff, radiation workers, patients and visitors. The present study was performed to check the areas with high radiation exposure during the running of the cyclotron.
**Materials and Methods:** Ionization chamber based detector and Geiger Muller counter were used to record the exposure levels before, during and after operating the cyclotron. The readings were recorded at various locations where a high radiation field was expected. The results were recorded, tabulated and analyzed.
**Results:** Highest exposure level (0.93μSv/hr) was found in the hot lab after the bombardment. The next highest exposure level was found near the vault wall (0.26μSv/hr) during production of the radioactivity in the cyclotron. The average exposure levels near the vaults door, console station, outer wall, back wall and adjacent rooms using the GM counter were 0.17, 0.16, 0.14, 0.18 and 0.14μSv/hr respectively. The average background levels recorded was 0.13μSv/hr.
**Conclusion:** The radiation survelliance around the self shelded cyclotron reflected that the radiation exposure levels outside the shield are minimal. The study showed no leakage of the radiation outside the vault. All the values are below recommended levels of exposure.
RS-7: Delay Tank- An overview in Indian regulatory context
**Subrata Pathak, Pankaj Tandon**
Radiological Safety Division, Atomic Energy Regulatory Board, Mumbai, India
**Introduction:** Delay tank is a tank or reservoir for the temporary holdup of radioactive waste coming out of a high dose radioiodine ward, to allow the activity to decay and bring down to the prescribed limit for discharge.
**Objective:** The requirement of Delay tank in a high-dose radioiodine ward where patients suffering from Ca-thyroid and neuroendocrine tumour are being treated with Radio-Iodine (RI) \[^131^I\] and Lu177 has been reviewed in light of present Indian regulatory context.
**Materials and Methods:** The clearance of RI from patients, its accumulation in the Delay tank and the amount of activity remaining in the Delay tank following physical decay has been simulated using simple mathematical model. Different case scenarios have been taken into account to achieve a better estimation of the cumulative activity in the Delay tank.
**Results and Conclusion:** This theoretical calculation confirms the requirement of Delay tank in high dose radioiodine facilities to meet the present regulatory limits for discharge of radioactivity into a public sewerage line.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Band structure engineering and low-loss propagation of information-carrying entities are the cornerstones of both electronics and magnonics. Whereas electronics exploits dissipative transfer of electric charges, magnonics is based on low-energy transport of angular momentum in the form of spin waves. Active control over the phase and the amplitude of spin waves offers promising prospects for nanotechnologies^[@CR1]^ because their wavelength is tunable down to the nanometer scale. Magnonic crystals provide this quintessential functionality^[@CR2]--[@CR7]^, with one-dimensional arrays of parallel magnetic stripes being an archetypical example^[@CR8]--[@CR13]^. Microwave magnetic fields excite collective spin-wave modes if the stripes are coupled via dipolar magnetic fields. As spin waves propagate through a magnonic crystal with positive group velocity when the magnetization aligns along the stripes, this Damon-Eshbach (DE) configuration is most relevant for magnonic devices. Patterning of ferromagnetic metals such as permalloy (Py), Co, and CoFeB into stripe arrays allows for versatile engineering of magnonic band structures^[@CR8]--[@CR15]^. For these materials, DE spin waves are highly dispersive and hence their group velocity is large. Strong dispersion also offers large tuning of the position and width of allowed minibands and forbidden frequency gaps by variation of the stripe width, lattice constant, or applied magnetic field. However, magnonic crystals made of ferromagnetic metals suffer from inefficient transmission because of strong Gilbert damping. The decay length of spin waves in these materials is limited to a few tens of micrometers^[@CR7]^.
Ferrimagnetic YIG exhibits magnetic damping about two orders of magnitude lower than 3d transition metals and is considered as the foremost material for magnonics^[@CR1],[@CR16]^. Research on YIG-based magnonic crystals is extensive, but in contrast to ferromagnetic metals, spin-wave propagation across individual YIG structures is hampered by weak dipolar coupling. This bottleneck, which stems from the small magnetization of YIG, limits the transmission of spin waves. Several alternatives have been investigated such as micrometer-thick YIG films with shallow air grooves^[@CR17]--[@CR21]^, metallic stripes on top^[@CR22],[@CR23]^, current carrying meander structures^[@CR24]^, width modulations^[@CR25]^, and light absorbers^[@CR26]^. Because spin-wave scattering in these magnonic crystals is weak compared to discrete lattices, many unit cells (*N* = 5--35 in refs. ^[@CR17]--[@CR26]^) are needed to open up bandgaps. The inefficiency of spin-wave reflection from single lattice units also restricts the bandgap size^[@CR27]^. In combination with the relatively flat spin-wave dispersion relation of YIG, this has thus far produced small bandgaps ranging from only a few to a few tens of MHz in YIG-based magnonic crystals.
Here, we experimentally demonstrate discrete YIG-based magnonic crystals combining robust tunable bandgaps of up to 200 MHz and low-loss transmission in allowed minibands over a distance of 200 μm. The crystals consist of 260-nm-thick YIG stripes that are separated from each other by only a few (*N* = 2--4) narrow channels of air or CoFeB. Using broadband spin-wave spectroscopy and micromagnetic simulations, we show how the forced precessions of magnetization in CoFeB efficiently enhance the coupling between quasi-confined spin-wave modes in YIG stripes, causing longer distance propagation of DE spin waves at allowed frequencies. In contrast, transmission of spin waves inside the bandgaps is almost completely suppressed. The positions and widths of the bandgaps are tunable by variation of the external bias field, lattice constant, or stripe width. In comparison, the band structure of YIG-based magnonic crystals without CoFeB (airgap-separated stripes) exhibits slightly larger bandgaps, but propagating spin waves damp out more quickly because of reduced dipolar coupling between the YIG stripes.
Results {#Sec2}
=======
Spin-wave transmission in continuous YIG films {#Sec3}
----------------------------------------------
Single-crystal ferrimagnetic YIG films with a thickness of 260 nm were grown on (111)-oriented Gd~3~Ga~5~O~12~ (GGG) substrates using pulsed laser deposition (PLD). After growth, the film structure was characterized by X-ray diffraction (XRD) and transmission electron microscopy (TEM). Figure [1a](#Fig1){ref-type="fig"} shows a cross-sectional high-resolution TEM image and the corresponding diffraction pattern of an as-grown YIG film on GGG. The well-ordered atomic structure and sharp diffraction spots illustrate coherent epitaxial growth of YIG(111) on GGG(111). Magnetic damping in the YIG films was measured using a vector network analyzer ferromagnetic resonance (VNA-FMR) setup. Figure [1b](#Fig1){ref-type="fig"} presents the linewidth of VNA-FMR spectra as a function of resonance frequency. From a linear fit to the data using Δ*f* = 2*αf* + *υ*~g~Δ*k*^[@CR28]^, where *υ*~g~ is the group velocity and Δ*k* is the excitation spectrum width of the antenna, we find a Gilbert damping constant of *α* = (4.4 ± 1.5) × 10^−4^. Spin-wave transmission in the YIG films was studied using broadband spin-wave spectroscopy. Figure [1c](#Fig1){ref-type="fig"} shows a schematic of the measurement geometry. Two 6-μm-wide antennas made of 3 nm Ta/200 nm Au were patterned on the YIG films with a separation of *x* = 200 μm. Spin waves were excited by spatially inhomogeneous microwave magnetic fields from the first antenna and inductively detected by the second antenna using a vector network analyzer. The spin waves had wave vectors ranging from zero to about *π*/*w*, where *w* = 6 μm corresponds to the antenna width^[@CR29]^. We characterized spin-wave transmission by recording the S~12~ scattering parameter.Fig. 1Properties of continuous nanometer-thick YIG films. **a** High-resolution TEM image and corresponding diffraction pattern of a 260-nm-thick YIG film. The dashed line indicates a sharp interface between the GGG(111) substrate and single-crystal YIG film. The white scale bar corresponds to 5 nm. **b** Linewidth of VNA-FMR spectra as a function of frequency. The data are obtained by variation of the external magnetic bias field. **c** Schematic of the broadband spin-wave spectroscopy measurement geometry. **d** Spin-wave transmission spectrum (imaginary part of S~12~) for a continuous 260-nm-thick YIG film at a magnetic bias field of 5 mT. **e** Spin-wave group velocity (*υ*~g~) and decay length (*l*~d~) in YIG. **f** Imaginary part of S~12~ at different external magnetic bias fields
Figure [1d](#Fig1){ref-type="fig"} shows a typical transmission spectrum (imaginary part of S~12~) of a 260-nm-thick YIG film at an external magnetic bias field of 5 mT, which suffices to fully saturate the magnetization. An oscillating spin-wave signal is measured above the FMR frequency of \~1.0 GHz. The strong oscillations indicate efficient spin-wave propagation between the two antennas. The group velocity of the spin waves (*υ*~g~) is extracted from the frequency separation (*δf*, indicated in Fig. [1d](#Fig1){ref-type="fig"}) using *υ*~g~ = *δf* × *x*^[@CR30],[@CR31]^. For frequencies ranging from 1.1 to 1.8 GHz, we find a gradual decrease of the group velocity from 11 to 5.5 km s^−1^ (upper panel in Fig. [1e](#Fig1){ref-type="fig"}). This method assumes the spin-wave dispersion relation to be linear. The small non-linearity of the actual dispersion curve produces an estimated error of 4% in the group velocity values. The lowering of *υ*~g~ with frequency is explained by a flattening of the DE spin-wave dispersion relation with increasing wave vector. The large group velocity in our 260-nm-thick YIG films enables transport of spin waves over long distances. Using *l*~d~ = *υ*~g~/2*παf*, we extract a spin-wave decay length (*l*~d~) of up to 4 mm in a small magnetic field of 5 mT (lower panel in Fig. [1e](#Fig1){ref-type="fig"}). This low-loss transmission provides a solid platform for the design of magnonic crystals. Figure [1f](#Fig1){ref-type="fig"} shows the imaginary part of S~12~ for different external magnetic bias fields. Asymmetric spin-wave intensities for negative and positive bias fields indicate a non-reciprocal effect, caused by dissimilar DE spin-wave propagation at the surface and interface of the YIG film^[@CR32]--[@CR35]^.
Spin-wave transmission in YIG-based magnonic crystals {#Sec4}
-----------------------------------------------------
Magnonic crystals were fabricated by photolithography and argon-ion milling of periodic grooves in 260-nm-thick YIG films. The grooves were centered between the two microwave antennas and extended all the way to the GGG substrate. In addition to airgap-separated YIG stripes, we also fabricated bicomponent magnonic crystals by filling the grooves with 260 nm CoFeB using magnetron sputtering. A large number of magnonic crystals with different combinations of groove width (*w*) and lattice constant (*a*) were patterned. Figure [2a](#Fig2){ref-type="fig"} shows a schematic of our magnonic crystal and measured height profiles of samples without (orange line) and with (green line) CoFeB in the grooves. During broadband spin-wave spectroscopy, spin waves are excited in the continuous part of the YIG film near the first antenna. These spin waves propagate perpendicular to the stripes while a small external bias field (*H*~ext~) establishes the DE configuration (Fig. [2a](#Fig2){ref-type="fig"}). Figure [2b](#Fig2){ref-type="fig"} shows the imaginary part (top panel) and amplitude (bottom panel) of the S~12~ scattering parameter as a function of frequency for a magnonic crystal with *N* = 4 CoFeB-filled grooves, *a* = 30 μm, and *w* = 2.5 μm (hereafter referred to as *a*30*w*2.5). The bias field in this measurement is 5 mT. Four allowed minibands and three forbidden frequency gaps of \>100 MHz are observed. Transmission of spin waves in the bandgaps is almost completely suppressed, as illustrated by a reduction of the signal amplitude to the background level measured at *f* \< 1 GHz. In the allowed minibands, spin waves propagate efficiently through the bicomponent magnonic crystal. The spin-wave signal of the CoFeB/YIG stripe array is only a factor \~3 smaller than that of an unpatterned YIG film (blue line in Fig. [2b](#Fig2){ref-type="fig"}). From the frequency separation between maxima and minima in the ImS~12~ signal near the center of the minibands, we derive a spin-wave group velocity ranging from 11 km s^−1^ in the first band (\~1.20 GHz) to 4.4 km s^−1^ in the fourth band (\~1.73 GHz). These group velocities are similar to those measured on a continuous YIG film (Fig. [1e](#Fig1){ref-type="fig"}).Fig. 2Spin-wave transmission in nanometer-thick YIG-based magnonic crystals. **a** Schematic of a one-dimensional magnonic crystal. The crystal consists of physically separated YIG stripes with a thickness of 260 nm and narrow air grooves or grooves that are filled by CoFeB. Two 6-μm-wide microwave antennas are patterned on top of YIG. Their separation is *x* = 200 μm. The bottom graph shows height profiles for magnonic crystals with air grooves (orange line) and grooves that are filled by CoFeB (green line). **b** Spin-wave transmission spectra (imaginary part and amplitude of S~12~) for a *a*30*w*2.5 CoFeB/YIG magnonic crystal at a magnetic bias field of 5 mT. The reference spectrum (blue line) is recorded on a continuous YIG film. **c** Imaginary part of S~12~ for the same crystal at different magnetic bias fields. **d** Dependence of spin-wave transmission spectra (amplitude of S~12~) on magnetic field angle in a *a*30*w*2.5 CoFeB/YIG crystal. In this graph, 90° corresponds to the excitation of DE spin waves. **e**, **f** S~12~ amplitude as a function of **e** lattice constant and **f** CoFeB stripe width. In **e**, *w* = 2.5 μm. In **f**, *a* = 50 μm. All the data are recorded on magnonic crystals with *N* = 4 grooves
Figure [2c](#Fig2){ref-type="fig"} shows spectra of the imaginary part of S~12~ at different external magnetic bias fields. In our magnonic crystal, large bandgaps persist up to tens of mT. The forbidden gaps are largest near zero bias field and gradually diminish with increasing field strength. We note that magnetic switching in the YIG and CoFeB stripes occurs independently at a field of about 0.1 and 2.0 mT, respectively (Supplementary Fig. [1](#MOESM1){ref-type="media"}). An antiparallel magnetization configuration is therefore attained in small bias fields. Judging by the continuous curves of Fig. [2c](#Fig2){ref-type="fig"}, switching to a parallel magnetization state at −2.0 or +2.0 mT does not affect the spin-wave transmission signal. Reprogramming of the magnonic band structure by toggling between differently aligned magnetization states, as demonstrated for discrete Py wires^[@CR11]^, does therefore not occur in our bicomponent magnonic crystals. We attribute this invariance to the large difference in stripe width and the relatively small precession of magnetization in CoFeB at 1.0--1.8 GHz. Figure [2d](#Fig2){ref-type="fig"} presents the angular dependence of spin-wave transmission for a 5 mT field. In this graph, a field angle of *ϕ*~H~ = 90° corresponds to the DE geometry (*M* \|\| stripes) and *ϕ*~H~ = 0° or 180° would represent results for backward volume magnetostatic spin waves (BVMSWs, *M* ⊥ stripes). Three bandgaps are resolved for angles between 60° and 120°. Rotation of the field away from the DE configuration reduces the signal intensity, bandgap positions, and gap widths (the angular dependence of the latter parameter is more clearly illustrated in Supplementary Fig. [2](#MOESM1){ref-type="media"} for a *a*50*w*5 crystal). These effects are explained by a weakening of the excitation efficiency and a flattening of the dispersion relation upon rotation of magnetization toward the spin-wave wave vector^[@CR31],[@CR36]^. We prepared a large number of bicomponent magnonic crystals by varying parameters *a* and *w* in 1 and 0.25 μm steps. Figure [2e, f](#Fig2){ref-type="fig"} summarize the transmission of spin waves in all these crystals. Using only four CoFeB stripes, we observe up to seven bandgaps and strong tuning of bandgap parameters with *a*. While the CoFeB stripe width affects the magnonic band structure less, the efficiency of spin-wave transmission reduces quickly when *w* becomes large.
To gauge the effect of CoFeB on the performance of our magnonic crystals in more detail, we directly compare spin-wave transmission spectra of bicomponent lattices to those of identical YIG stripe arrays without CoFeB. Figure [3](#Fig3){ref-type="fig"} presents results for three different magnonic crystals, *a*30*w*2.5, *a*30*w*5, and *a*50*w*5, with orange and green lines representing data for pure YIG and CoFeB/YIG lattices, respectively. Whereas both types of crystals show robust bandgaps, their widths are slightly larger for stripe arrays with air grooves. Moreover, the spin-wave signal decays faster in crystals made of pure YIG, which is particularly apparent for *w* = 5 μm. Filling of the air grooves with CoFeB thus enhances the efficiency of spin-wave transmission in nanometer-thick YIG-based magnonic crystals.Fig. 3Spin-wave dispersion relations and Bragg reflection. **a**--**c** Transmission spectra of CoFeB/YIG bicomponent magnonic crystals (green lines) and YIG stripes with airgaps (orange lines). The magnonic crystals consist of *N* = 4 grooves. The lattice constants and groove widths are indicated in the graphs. **d**--**f** Calculated dispersion relations of DE spin waves for the bicomponent magnonic crystals. The vertical and horizontal dashed lines show the center frequencies of bandgaps and corresponding wave vectors
We quantify the bandgap size and efficiency of spin-wave transmission by depicting their dependence on parameter *w* in Fig. [4](#Fig4){ref-type="fig"}. The spin-wave transmission spectra of Fig. [4a--c](#Fig4){ref-type="fig"} compare results for pure YIG (orange lines) and CoFeB/YIG (green lines) lattices with *a* = 30 μm and *w* = 2, 4, and 6 μm. The blue lines in these graphs are measured on a continuous area of the same YIG film. From these data (and results on lattices with *w* = 3 and 5 μm), we extracted the spin-wave transmission signal in the first three minibands relative to the S~12~ amplitude measured on continuous YIG (Fig. [4d](#Fig4){ref-type="fig"}). The results demonstrate that the efficiency of spin-wave transmission drops 16 and 22 times in the 2nd and 3rd miniband if the air-groove width is enlarged from 2 to 6 μm. In contrast, the S~12~ signal only reduces by a factor \<3 upon an increase of *w* if the grooves are filled with CoFeB. Because of these dissimilar dependencies, the transmission of spin waves in higher order minibands of CoFeB/YIG lattices is at least one order of magnitude larger than in crystals made of pure YIG if *w* \> 5 μm. Figure [4e](#Fig4){ref-type="fig"} shows the dependence of bandgap size on groove width. The bandgap increase slightly with *w*, but the difference between YIG and CoFeB/YIG magnonic crystals remains small (\<15%).Fig. 4Dependence of spin-wave transmission and bandgap size on groove width. **a**--**c** Transmission spectra of pure YIG (orange lines) and CoFeB/YIG (green lines) magnonic crystals with *N* = 4 grooves. The blue line depicts a reference spectrum measured on a continuous part of the same YIG film. **d** Dependence of spin-wave transmission on groove width *w* in the first three minibands. The amplitudes of the S~12~ scattering parameter in the magnonic crystals (Amp. S~12~\_MC) are compared to those recorded on continuous YIG (Amp. S~12~\_YIG film). **e** Variation of the bandgap size with groove width
The center frequencies of the bandgaps are accurately mapped onto the spin-wave dispersion relations of the magnonic crystals at wave vectors that satisfy conditions of Bragg reflection. To illustrate this, we plot the dispersion curves for DE spin waves in *a*30*w*2.5, *a*30*w*5, and *a*50*w*5 CoFeB/YIG magnonic crystals in the lower panels of Fig. [3](#Fig3){ref-type="fig"}. The dispersion relations were calculated using^[@CR37]^$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$f = \frac{{\gamma \mu _0}}{{2\pi }}\sqrt {\left( {H_{{\mathrm{ext}}} + H_{{\mathrm{ani}}}} \right)\left( {H_{{\mathrm{ext}}} + H_{{\mathrm{ani}}} + M_{\mathrm{s}}} \right) + \frac{{M_{\mathrm{s}}^2}}{4}\left( {1 - e^{ - 2kd}} \right)} .$$\end{document}$$
As input parameters, we used $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\textstyle{\gamma \over {2\pi }}}$$\end{document}$ = 28 GHz T^−1^, *d* = 260 nm, *μ*~0~*H*~ext~ = 5 mT, magnetic anisotropy field *μ*~0~*H*~ani~ = 1.5 mT, and effective saturation magnetizations *M*~s~ = 185 kA m^−1^ (for *a*30*w*2.5 and *a*50*w*5) and *M*~s~ = 192 kA m^−1^ (for *a*30*w*5). The values of *μ*~0~*H*~ani~ and *M*~s~ were extracted from the dependence of FMR frequency on external magnetic field (Supplementary Fig. [3](#MOESM1){ref-type="media"}). These measurement data were obtained by recording the S~12~ scattering parameter in transmission on the same magnonic crystals. The slight increase of *M*~s~ for the *a*30*w*5 crystal is attributed to the larger CoFeB/YIG ratio^[@CR14],[@CR38]^. For all crystals, the center frequencies of the bandgaps correspond to wave vectors *nπ*/*a*, where *n* is the order number of the bandgap and *π*/*a* is the first Brillouin zone boundary. Efficient Bragg reflection of incoming spin waves in our magnonic crystals is thus responsible for the opening of robust bandgaps with high rejection efficiencies.
We note the occurrence of an extra feature at \~1.2 GHz in the lowest frequency pass band of the *a*30*w*4, *a*30*w*5, and *a*30*w*6 crystals (Figs. [3b](#Fig3){ref-type="fig"} and [4b, c](#Fig4){ref-type="fig"}). This feature, which signifies a real suppression of the spin-wave transmission signal, appears more clearly if the grooves become wider. Moreover, the suppression is stronger if the YIG stripes are separated by air rather than CoFeB. At the feature frequency, the wavelength of excited spin waves matches the size of the magnonic crystal (i.e., the distance between the first and last groove). For more details see Supplementary Fig. [4](#MOESM1){ref-type="media"}.
The transmission spectra presented thus far are recorded on magnonic crystals with *N* = 4 grooves. This limited number already suffices for the formation of deep bandgaps. In previously studied magnonic crystals, a larger number of spin-wave reflectors is often required to significantly suppress the transmission of spin waves in the frequency gaps^[@CR5]--[@CR7]^. To test whether the number of grooves in our nanometer-thick YIG-based magnonic crystals can be reduced even further, we fabricated samples with *N* = 2, 3 .. 6. Figure [5](#Fig5){ref-type="fig"} shows results for CoFeB/YIG samples with *a* = 30 μm and *w* = 2.5 μm. The experiments demonstrate that the magnonic crystal with only two grooves already exhibits pronounced bandgaps. With increasing *N*, the bandgaps deepen and the efficiency of spin-wave transmission in the allowed minibands decreases. The width of the bandgap does not depend strongly on *N*, which is in line with previous observations that this parameter is mainly defined by the reflection efficiency of a single reflector^[@CR27]^.Fig. 5Dependence of spin-wave transmission spectra on the number of grooves. Transmission spectra of CoFeB/YIG bicomponent magnonic crystals with *w* = 2.5 μm, *a* = 30 μm, and different numbers of CoFeB-filled grooves (*N*). For clarity, the spectra are plotted with constant offsets
Simulations of spin-wave profiles and transmission spectra {#Sec5}
----------------------------------------------------------
We used the MuMax3 micromagnetic solver^[@CR39]^ to simulate spin waves in YIG and CoFeB/YIG magnonic crystals. The simulation area consisted of a 655-μm-long and 260-nm-thick YIG film with *N* periodic airgaps or CoFeB stripes (Fig. [6a](#Fig6){ref-type="fig"}). The structure was discretized using finite-difference cells of size *x* = 40 nm, *y* = 40 nm and *z* = 32.5 nm. One-dimensional periodic boundary conditions were applied along the *y*-axis to mimic a lattice with infinitely long stripes. To reduce reflections from the edges along *x*, we added two 5-μm-wide areas with a high Gilbert damping constant of *α* = 0.5. As input parameters, we used *M*~s~ = 185 kA m^−1^ (YIG), *M*~s~ = 1150 kA m^−1^ (CoFeB), *A*~ex~ = 3.1 pJ m^−1^ (YIG), *A*~ex~ = 16 pJ m^−1^ (CoFeB), *α* = 0.003 (YIG) and *α* = 0.005 (CoFeB). For YIG, the relatively large damping parameter was intentionally selected to limit the computation time and reduce reflections from the edges of the simulation area. Moreover, we added a 0.16-μm-wide non-magnetic layer between the YIG and CoFeB stripes. During the course of our simulation study, we noted that good agreements with experimental spin-wave transmission spectra could only be attained with the inclusion of a thin non-magnetic spacer layer. We attribute the formation of this layer to argon-ion milling in our lithography process. Complex oxides are particularly sensitive to ion-beam milling and, therefore, we anticipate YIG to be structurally and magnetically modified at the groove edges. Supplementary Figure [5](#MOESM1){ref-type="media"} shows how the thickness of the non-magnetic layer influences the transmission spectra. We used sinc-function-type field pulses with a cutoff frequency of 3 GHz to generate spin-wave intensity maps and sinusoidal ac magnetic fields with fixed frequencies to determine the time evolution of spin-wave profiles. The fields were applied along the *x*-axis over a 6-μm-wide area in the continuous YIG part. To visualize the propagation of spin waves in magnonic crystals, we analyzed the *x*-component of magnetization (*m*~x~), whose dynamic response is much stronger than *m*~z~ because of the strong demagnetization field in our thin-film geometry.Fig. 6Micromagnetic simulations. **a** Spatial distribution of spin-wave intensity in a *a*30*w*2.5 bicomponent magnonic crystal with *N* = 4 CoFeB stripes. The right panel compares simulated (solid green line) and measured (dashed blue line) transmission spectra. The inset illustrates the simulation geometry. **b**, **c** Time evolution of spin-wave profiles during ac excitation with a sinusoidal magnetic field at a fixed frequency of **b** 1.44 GHz and **c** 1.35 GHz. Spin-wave profiles after an excitation time of 30 ns are shown on the right side of the graphs. The system reached steady-state excitation at this time. **d**--**f** Similar data as in **a**--**c**, but for a *a*30*w*2.5 crystal without CoFeB, i.e., discrete YIG stripes separated by airgaps. **g** Amplitude of *m*~x~ near the first groove (top panel). The bold blue and green lines correspond to data recorded in the boxes of **b**, **e**. The curves show data for a fixed groove width of *w* = 2.5 μm, but different amounts of CoFeB (*L*~CoFeB~). The bottom panel of **g** shows the amplitude of the *x*-component of the dynamic dipolar field for different amounts of CoFeB. **h** Variation of the dynamic dipolar field near the edge of the first groove (*x* = 62.9 μm in **g**) as a function of *L*~CoFeB~ for *a*30*w*2.5 and *a*30*w*5 magnonic crystals
Figure [6a](#Fig6){ref-type="fig"} shows the spatial distribution of spin-wave intensity in a bicomponent *a*30*w*2.5 magnonic crystal. In this simulation, spin waves with frequencies ranging from 1.0 to 2.5 GHz are excited near *x* = 0 μm. Propagating spin waves with a frequency of \~1.35 GHz and wave vectors *k* ≈ *π*/*a* are efficiently reflected by the one-dimensional stripe array. Consequently, the spin-wave intensity in subsequent YIG stripes drops quickly and transmission across the entire magnonic crystal is almost completely suppressed. Bandgaps are formed also when conditions for higher-order Bragg reflections are met (*k* = *nπ*/*a*). The positions and widths of the simulated bandgaps correspond closely to our experimental data, as illustrated by the transmission spectra on the right side of Fig. [6a](#Fig6){ref-type="fig"}. The simulated spectrum (green solid line) was obtained by averaging *m*~x~ over a 6-μm-wide area at *x* = 200 μm. Apart from a stronger decay of spin waves in the simulations (caused by the use of *α* = 0.003 for YIG), the bandgap parameters agree well. In the allowed minibands between the frequency gaps, spin waves efficiently propagate through the bicomponent magnonic crystal. Nodes in the spin-wave intensity map of Fig. [6a](#Fig6){ref-type="fig"} demonstrate that the spin waves are partially confined in the YIG stripes. These quasi-standing modes couple to each other via dynamic dipolar fields. As the dynamic coupling field between neighboring stripes decreases monotonically with node number^[@CR5],[@CR10]^, the transmission signal is largest in the first miniband.
Figure [6b](#Fig6){ref-type="fig"} depicts the time evolution of spin-wave transmission through a *a*30*w*2.5 CoFeB/YIG lattice at a frequency of 1.44 GHz, which corresponds to the second miniband center. After the onset of continuous excitation at *t* = 0 s, spin waves propagate toward the magnonic crystal. Dynamic dipolar coupling between the YIG stripes results in an efficient perpetuation of the excitation across the crystal. The phase of spin waves in neighboring YIG stripes is preserved, illustrating the collective character of the modes. In Fig. [6c](#Fig6){ref-type="fig"}, we show a similar simulation at 1.35 GHz. Because this frequency is located in the first bandgap, spin waves are effectively reflected by the lattice and their propagation is limited to 1--2 YIG stripes.
For comparison, we now present simulations for the same *a*30*w*2.5 crystal, but with CoFeB being replaced by air. Figure [6d--f](#Fig6){ref-type="fig"} summarizes the results. In analogy to the bicomponent crystal, several allowed minibands and bandgaps are simulated. Based on these data, we conclude that the separating material (CoFeB versus air) has two main effects. First, bandgaps of the pure YIG magnonic crystal are a bit wider than those of the CoFeB/YIG lattice. Second, transmission of spin waves is more efficient if the air grooves are filled by CoFeB. Both observations agree with the experimental data of Figs. [3](#Fig3){ref-type="fig"} and [4](#Fig4){ref-type="fig"}.
The simulations in Fig. [6b, c](#Fig6){ref-type="fig"} indicate that magnetization precessions in the narrow CoFeB stripes are much smaller than in YIG. To quantify the effect of CoFeB on dynamic coupling between the YIG stripes, we performed micromagnetic simulations on *a*30*w*2.5 and *a*30*w*5 crystals (Fig. [6g, h](#Fig6){ref-type="fig"}) with varying amounts of CoFeB. Starting from a pure airgap of fixed width, we positioned narrow CoFeB stripes in the groove center. The width of CoFeB stripe (*L*~CoFeB~) was varied in steps from 0 (pure airgap) to *w* (groove fully filled with CoFeB). Figure [6g](#Fig6){ref-type="fig"} summarizes the magnetization precession amplitude and *x*-component of the dynamic dipolar field near the first groove of the *a*30*w*2.5 crystal. The data are simulated at a frequency that corresponds to the center of the second allowed miniband. Despite the relative small value of *m*~x~ in the CoFeB stripe, the use of CoFeB substantially enhances the dynamic dipolar field. As a result, the YIG stripes are coupled more strongly and the efficiency of spin-wave transmission is enhanced compared to the same magnonic crystal with airgaps. Figure [6h](#Fig6){ref-type="fig"} summarizes the dependence of the dynamic dipolar field on the amount of CoFeB in the first groove of *a*30*w*2.5 and *a*30*w*5 lattices.
Finally, we describe simulations on spin-wave transmission in magnonic crystals with only *N* = 2 grooves. Figure [7](#Fig7){ref-type="fig"} compares results for *a*30*w*2.5 crystals with CoFeB stripes and airgaps. The simulations confirm that only two spin-wave reflectors suffice for the formation of bandgaps and allowed minibands in discrete magnonic crystals of nanometer-thick YIG, in agreement with experimental data (Fig. [5](#Fig5){ref-type="fig"}). Again, the transmission of spin-waves in the allowed minibands is larger for the bicomponent CoFeB/YIG lattice, while air grooves produce slightly more pronounced bandgaps.Fig. 7Simulations for discrete magnonic crystals with only two units. **a**, **b** Spatial distribution of spin-wave intensity in a *a*30*w*2.5 crystal with **a** *N* = 2 CoFeB stripes and **b** *N* = 2 airgaps. The right panels show simulated transmission spectra (solid green lines) and an experimental measurement (dashed blue line) for the bicomponent lattice
Discussion {#Sec6}
==========
Our results demonstrate that robust bandgaps and high transmission efficiency can be achieved in magnonic crystals based on discrete YIG stripes. YIG films with patterned air grooves have been considered before^[@CR17]--[@CR21]^. In these studies, the thickness of YIG ranged from a few to several tens of micrometers and grooves did not extend throughout the film. Besides technical challenges of patterning deep narrow grooves in thick YIG films, a strong increase of transmission losses with groove depth justifies the omission of stripe arrays on this thickness scale^[@CR17],[@CR18]^. For micrometer-thick YIG stripes, spin-wave propagation is hampered because in-plane (*m*~x~) and out-of-plane (*m*~z~) magnetization precessions produce comparable dipolar coupling fields. The combination of these dynamic fields, with one favoring in-phase and the other forcing out-of-phase oscillations, suppresses the spin-wave signal. In our nanometer-thick YIG-based magnonic crystals, the situation is different. Here, the demagnetization field reduces *m*~z~ and, hence, stripes predominantly interact via in-plane dynamic dipolar fields. Downscaling the thickness of YIG-based magnonic crystals, which is now possible because of recent advancements in high-quality YIG film growth^[@CR40]--[@CR44]^, thus provides new opportunities for low-loss manipulation of spin waves in magnonic devices.
Besides the positive impact of thickness reduction, our experimental and simulation data show that spin-wave propagation improves further if the air grooves between YIG stripes are filled by CoFeB. We explain this effect by an enhancement of dynamic dipolar coupling via forced magnetization precessions in the narrow CoFeB stripes (Fig. [6g, h](#Fig6){ref-type="fig"}). Despite their small amplitude, these precessions produce a significant dipolar field because the magnetization of CoFeB is large (*M*~s,CoFeB~ ≈ 8*M*~s,YIG~). The dynamic field that the CoFeB stripes produce adds to the dipolar field from quasi-standing modes in the YIG stripes and, consequently, it promotes spin-wave propagation across the magnonic crystal. Enhanced stripe-to-stripe coupling also affects the band structure. Near-perfect confinement of spin-wave modes in the stripes of a one-dimensional crystal with parameters *w* (gap width) and *a* (lattice constant) would in theory produce narrow minibands for spin waves with *k* = *nπ*/(*a* − *w*) and these bands would be separated by large frequency gaps. Dynamic dipolar coupling between stripes reduces the confinement of spin-wave modes. For constant *w* and *a*, an increase of the coupling strength would thus enhance the miniband width and reduce the bandgap size. This is exactly what we observe in our experiments (Figs. [3](#Fig3){ref-type="fig"} and [4](#Fig4){ref-type="fig"}) and simulations (Figs. [6](#Fig6){ref-type="fig"} and [7](#Fig7){ref-type="fig"}) when the air grooves of our YIG-based magnonic crystals are filled by CoFeB.
The opening of clear bandgaps down to *N* = 2 in our fully discrete nanometer-thick YIG-based magnonic crystals is unique. To illustrate the effect of groove depth on the magnonic band structure, we performed micromagnetic simulations on lattices with *N* = 2 air grooves that are 12.5 or 50% deep (Supplementary Fig. [6a](#MOESM1){ref-type="media"}). In both cases, the two grooves do not open up bandgaps in the spin-wave transmission spectra. Bandgaps with only two scatterers are attained if spin waves are efficiently reflected by grooves that fully extend throughout the YIG film. Compared to magnonic crystals made of ferromagnetic metals, our YIG-based magnonic crystals have the advantage of much lower magnetic damping. Because of this, multiple reflections on periodic scatterers are more readily attained, producing stronger intensity modulations in spin-wave transmission spectra. Simulations on crystals with *N* = 2 and damping parameters *α* = 0.003 or *α* = 0.01 (Supplementary Fig. [6b](#MOESM1){ref-type="media"}) demonstrate that low magnetic damping is essential for versatile band structure engineering in lattices with small *N*.
Finally, we discuss tuning of the magnonic band structure. Figure [8](#Fig8){ref-type="fig"} summarizes how parameters of the first two bandgaps depend on external magnetic bias field, lattice constant, and CoFeB stripe width. Large tuning of the bandgap frequency (1.2 GHz → 3.4 GHz) and gap size (140 MHz → 50 MHz) is measured and simulated for a *a*30*w*2.5 crystal if the bias field is increased from 3 to 45 mT (Fig. [8a, b](#Fig8){ref-type="fig"}). The monotonic decrease of the bandgap width is explained by a flattening of the spin-wave dispersion curve at large magnetic fields^[@CR23]^. Geometrical parameters affect the band structure via a change of the Brillouin zone boundary (*π*/*a*) or shift in the spin-wave dispersion relation. The latter effect originates from a dependence of the effective magnetization on CoFeB/YIG ratio. Because spin-wave modes are located mostly in the YIG stripes, the frequency and size of the bandgap vary little with CoFeB stripe width (Fig. [8e, f](#Fig8){ref-type="fig"}). The band structure is affected more by the lattice constant. A shift of the Brillouin zone boundary to larger wave vectors enhances the frequency and size of the bandgap for small *a* (Fig. [8c, d](#Fig8){ref-type="fig"})). Based on the parameter space depicted in Fig. [8](#Fig8){ref-type="fig"}, we conclude that nanometer-thick YIG-based magnonic crystals allow for large tunable bandgaps of 50--200 MHz.Fig. 8Magnonic band structure tuning. **a**--**f** Measured (empty symbols) and simulated (solid symbols) positions and widths of the first (circles) and second (squares) bandgaps of bicomponent CoFeB/YIG magnonic crystals as a function of **a**, **b** magnetic bias field, **c**, **d** lattice constant, and **e**, **f** CoFeB stripe width
In summary, we have demonstrated operation of fully discrete YIG-based magnonic crystals for the first time. Low-loss transmission in combination with robust spin-wave manipulation is achieved by limiting the thickness of YIG stripes to the nanometer scale. Because the demagnetization field in our thin crystals suppress magnetization precession in the out-of-plane direction, the YIG stripes effectively interact via dynamic dipolar fields in the film plane. Dynamic magnetic coupling between the YIG stripes is further enhanced if the air grooves that separate them are filled by CoFeB. Incorporation of CoFeB increases the transmission of spin waves in allowed minibands and slightly reduces the size and depth of the bandgaps. The combination of efficient Bragg scattering on airgaps or narrow CoFeB stripes and low magnetic damping in YIG already produces clear bandgaps for the absolute minimum of two scattering units. Bandgap sizes are tunable from 50 to 200 MHz by variation of the lattice constant and groove width. Our magnonic crystals offer large versatility in the design of magnonic band structures and provide a promising avenue towards the realization of energy-efficient magnonic devices.
Methods {#Sec7}
=======
Sample fabrication {#Sec8}
------------------
We grew YIG films with a thickness of 260 nm on GGG(111) substrates using PLD. The GGG substrates were ultrasonically cleaned in acetone and isopropanol before loading into the deposition chamber. We degassed the substrates at 550 °C for 15 min. After this, oxygen was inserted into the chamber and the temperature was raised to 800 °C at a rate of 5 °C per minute. YIG films were deposited from a stoichiometric target in an oxygen partial pressure of 0.13 mbar. We used an excimer laser with a pulse repetition rate of 2 Hz and a laser fluence of 1.8 J cm^−2^. Following film growth, we first annealed the YIG films at 730 °C for 10 min in an oxygen environment of 13 mbar and then cooled them down to room temperature at a rate of −3 °C per minute. The deposition process resulted in single-crystal YIG films, as confirmed by XRD and TEM measurements. The magnonic crystals were fabricated by photolithography. Using a Laserwriter LW405 system, we first defined periodic one-dimensional stripes in a resist layer on top of the 260-nm-thick YIG films. Next, we used argon-ion milling to pattern deep grooves into the YIG film until the GGG substrate was reached. For bicomponent magnonic crystals, this step was followed by deposition of 260 nm Co~40~Fe~40~B~20~ using magnetron sputtering at room temperature. Finally, we performed lift-off by placing the samples in a bath of acetone. For spin-wave characterization, parallel microwave antennas with a separation of 200 μm and width of 6 μm were patterned on top of the YIG films. The antennas were fabricated by a similar laser-writing process and consisted of 3 nm Ta and 200 nm Au. Both metals were deposited by magnetron sputtering.
Broadband spin-wave spectroscopy {#Sec9}
--------------------------------
The setup for spin-wave characterization consisted of a two-port VNA (Agilent N5222A) and a home-built electromagnet probing station. We recorded spin-wave transmission spectra by measuring the S~12~ scattering parameter. The excitation power of the microwave signal was set at −10 dBm to avoid nonlinear excitations. We also measured transmission spectra with an excitation power of −30 dBm and, except for stronger background noise, we found identical spin-wave characteristics. We used a frequency sweep method to record spectra for different magnetic bias fields. To improve contrast, we subtracted a reference spectrum taken at 100 mT from imaginary S~12~ data (Figs. [1d, f](#Fig1){ref-type="fig"} and [2b, c](#Fig2){ref-type="fig"}). The amplitudes of S~12~ shown in Figs. [2](#Fig2){ref-type="fig"}b, d--f, [3](#Fig3){ref-type="fig"}, [4a--c](#Fig4){ref-type="fig"} and [5](#Fig5){ref-type="fig"} are raw data.
Micromagnetic simulations {#Sec10}
-------------------------
We performed micromagnetic simulations using open-source GPU-accelerated MuMax3 software. The dimensions of the YIG film were set to *x* = 655 μm, *y* = 1.5 μm, and *z* = 260 nm and one-dimensional periodic boundary conditions were applied along the *y*-axis. We discretized the simulations in 40 × 40 × 32.5 nm^3^ cells. To mimic the experimental magnonic crystals and measurement geometry, we placed *N* CoFeB stripes or airgaps between 6-μm-wide excitation and detection areas. As input parameters, we used experimentally derived parameters for YIG and CoFeB. An external magnetic bias field was applied along the *y*-axis. For excitation, we used sinc-function-type magnetic field pulses with a cut-off frequency of 3 GHz or sinusoidal ac magnetic fields. In both cases, the field strength was 0.5 mT. During continuous excitation, the time evolution of the *x*-component of magnetization (*m*~x~) was recorded for 150 ns. Spatially resolved spin-wave intensity maps were obtained by Fourier-transforming the evolution of *m*~x~ on a cell-by-cell basis.
Supplementary information
=========================
{#Sec11}
Supplementary Information Peer Review File
**Publisher's note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
=================================
**Supplementary Information** accompanies this paper at 10.1038/s41467-018-07893-5.
This work was supported by the European Research Council (Grant Nos. ERC-2012-StG 307502-E-CONTROL and ERC-PoC-2018 812841-POWERSPIN) and the Academy of Finland (Grant Nos. 317918 and 316857). S.J.H. acknowledges financial support from the Väisälä Foundation. TEM characterization was conducted at the Aalto University Nanomicroscopy Centre (Aalto-NMC) and lithography was performed at the Micronova Nanofabrication Centre, supported by Aalto University. We also acknowledge the computational resources provided by the Aalto Science-IT project.
H.Q., G.-J.B., and S.v.D. initiated and designed the research. H.Q. fabricated the samples. H.Q. and G.-J.B. conducted the spin-wave measurements. L.Y. carried out TEM characterization. H.Q. and S.J.H. performed the micromagnetic simulations. S.v.D. supervised the project. H.Q. and S.v.D. wrote the manuscript. All authors discussed the results.
The data that support the findings of this study are available from the corresponding author upon request.
Competing interests {#FPar1}
===================
The authors declare no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Understanding the factors that set species geographical borders is a central issue in ecology and evolutionary biology [@pone.0092177-Gaston1] that has received increased attention in the face of the rapid distributional shifts associated with climate change. All species have limits to their geographical ranges beyond which they are not found, presumably because at these locations their tolerances and capacities are constrained or dispersal from source populations is limited [@pone.0092177-Bridle1], but these assumptions have rarely been tested [@pone.0092177-Sexton1]. The geographical ranges of species distributions may result from the interplay between many factors including habitat availability, local environmental conditions, metapopulation dynamics, biological interactions and species-specific traits, such as physiological tolerances, dispersal potential and vital rates [@pone.0092177-Case1], [@pone.0092177-Holt1].
Variations in population structure and dynamics across the species distributional ranges must reflect geographical variation in the dynamics of vital rates [@pone.0092177-Keitt1], [@pone.0092177-Purves1]. Therefore, to understand the determinants of species ranges it is fundamental to compare the vital rates at range boundaries to those of central populations and to understand how spatial variation in fitness translates into population-level differences in abundance [@pone.0092177-Sexton1]. Such comparisons are surprisingly rare, and even less frequent for southern latitude margins that are more susceptible to change as a consequence of climatic change. Exceptions include empirical comparisons between demographic traits at marginal and core locations for plants [@pone.0092177-Purves1], [@pone.0092177-Angert1], and marine organisms [@pone.0092177-Svensson1], [@pone.0092177-Lester1]. Studies on seaweeds, compared isolated [@pone.0092177-Viejo1] and integrated [@pone.0092177-Arajo1] demographic traits at central and southern range locations, or developed demographic models for central and marginal populations. However these studies did not include the most marginal populations [@pone.0092177-Svensson2]. Understanding the basis of the contrasting dynamics of edge and central populations will increase the reliability of models on the current, likely threatened e.g. [@pone.0092177-Diekmann1], [@pone.0092177-Nicastro1], status of populations at range margins and to predict their long-term response to current and future environmental conditions.
Geographically isolated populations of dioecious species with short direct dispersal are more likely to be affected by disturbance, especially if they are at the margins of their distributional range. Among this group of species are the two canopy forming fucoid seaweeds *Ascophyllum nodosum* and *Fucus serratus* that occur on sheltered intertidal shores in the north Atlantic. These are two ecologically and phylogenetically close foundation species with very different life-spans and phylogeographical histories. *A. nodosum* is a long-lived species able to produce aggregated clonal individuals by vegetative sprouting from its holdfast (base), with slow growth rate and an estimated maximum life span of a genetical individual of hundreds of years [@pone.0092177-berg1], whereas *F. serratus* does not produce clonal copies from the holdfast and has a shorter life span of about 3 to 4 years [@pone.0092177-Arrontes1]. Patterns of genetic diversity at the southern range edge are also very different for both species [@pone.0092177-Coyer1], [@pone.0092177-Hoarau1], [@pone.0092177-Olsen1]. Lower genetic diversity and high between population differentiation have been reported for *F. serratus* populations of the Iberian Peninsula when compared with central ones [@pone.0092177-Coyer1], [@pone.0092177-Hoarau1] which might limit the evolutionary potential of such marginal populations [@pone.0092177-Pearson1]. Similar genetic diversity and very low differentiation were found across the entire range of distribution of *A. nodosum* in the Atlantic, with exception of Brittany where higher values were reported [@pone.0092177-Olsen1]. This shows that genetic diversity is not reduced nor distinct at the southern edge. Population structure and demographic variables of both species at marginal and central locations are also different. *A. nodosum* marginal populations exhibited differentiated phenotypic traits in relation to central populations with higher values of reproductive output and densities and lower individuals' sizes, while *F. serratus* core and border populations did not show differences in vital rates [@pone.0092177-Arajo1].
The aim of the present study is to compare the demography of marginal and central populations of *A. nodosum* and *F. serratus* using stochastic matrix population models. These models are powerful tools to investigate the demography of stage-structured populations and allow the incorporation of temporal and spatial sources of variability in population dynamics [@pone.0092177-Caswell1]. Environmental stochasticity affects population dynamics and evolution of life-history traits and thus stochastic matrix models that include variations in vital rates due to temporal random fluctuations in the environment, were used in this study. Stochastic stage-based demographic models for the two species are developed and analyzed with focus on stochastic population growth rate, mean stable stage distribution, as well as specific sensitivity and elasticity of the stochastic population growth rate to changes in stage-specific vital rates. More specifically, we hypothesize that the stochastic population growth rate should be lower and more variable in the marginal than in the central populations and that the sensitivity of the stochastic population growth rate should be more sensitive to changes in fertility in the marginal populations. Furthermore, these analyses aim to contribute to global understanding of demographic contrasts in marginal versus central populations and their consequences for persistence of populations at their low-latitude distributional limits.
Materials and Methods {#s2}
=====================
Study Site {#s2a}
----------
This study was conducted on the northern coast of Portugal and the northern coast of Brittany in France covering, respectively, the southern-edge and central populations of *F. serratus* and *A. nodosum*. These species have maintained stable isolated southern-edge populations on the northern coast of Portugal for at least the last 50 years [@pone.0092177-Arajo2]. In Portugal, the study of *F. serratus* was performed in Viana do Castelo (41°41′27″N; 8°50′57″W) and Amorosa (41°38′26″N; 8°49′22″W) and for *A. nodosum* only in Viana do Castelo, the southernmost population in the eastern Atlantic (41°41′49″N; 8°51′08″W). In the French coast, two populations of both species were sampled, in Roscoff (48°43′38″N; 3°59′22″W) and in Santec (48°42′46″N; 4°01′23″W).
The central populations of both species were located in sheltered bays, protected by groups of coastal islands while the southern edge populations are on an open coast, exposed to prevailing NW swell. However, the area of the shore where *A. nodosum* and *F. serratus* occur is protected from the direct action of waves by a large extension of rocks located in front of their areas. This is demonstrated by the prevalence, in these sheltered zones of the shore, of brown algae assemblages dominated by *Pelvetia canaliculata*, *F. spiralis* and *F. vesiculosus* and *A. nodosum*.
No specific permissions were required to perform this study since it did not involve collection of specimens and none of the species studied had endangered or protected status.
All sites were strongly dominated by fucoid canopy at mid-intertidal sheltered shores. *A. nodosum* was found in the mid-high intertidal zone and *F. serratus* in the mid-low levels. The maximum tidal amplitude is approximately 4.0 m on the northern Portuguese coast and 9.0 m in the northern part of Brittany. Mean annual temperatures recorded during the study period under the *A. nodosum* canopy were 16°C in Portugal and 11°C in France [@pone.0092177-Arajo1].
Sampling Procedure {#s2b}
------------------
For each location (Portugal and France), at each site (Viana do Castelo and Amorosa in Portugal, and Roscoff and Santec in France) and for each species, three 1×1 m quadrats were randomly selected and marked in the corners with plastic tags and screws. However, for *A. nodosum* in Portugal only Viana do Castelo was sampled since it is the single southern border population that exists. The nearest known population occurs 10 s of kms further north, in Spain. For this reason, twelve 1×1 m plots were randomly spread along the 2000×30 m total distribution of the population in Viana do Castelo. Inside each quadrat and at each sampling occasion all the individuals were mapped. Individuals with a size large enough to hold a plastic cable tie or a glass bead tag were tagged. The basic information needed for the size structured matrix model is the fate of individuals, that is, if they survive or not. If they survive, then the change in size is important in order to calculate the probability of transition of individuals from one size class to another. Further, the number of recruits is needed to estimate the fertility. For that reason all new recruits within each quadrat were registered at each sampling occasion. To estimate individual size, the maximum length (l) and circumference (c) (in cm) of each individual were measured. The predictive equations previously obtained for each species at each location [@pone.0092177-Arajo1], for the relationship between algal volume \[length of the longest frond × maximum circumference squared (lc^2^)\] and dry weight, were used to calculate the dry biomass (in g) of all the individuals measured in the field. Sampling was done from 2005 to 2009 at variable dates, depending on the knowledge available about each species and each location prior to the start of the experiment. For *A. nodosum* in France, based on previous personal observations about the occurrence of a single reproductive period yearly, sampling was done once a year from 2006 to 2009 (summer). Because similar information was not available for *F. serratus* in France, sampling was done twice a year during the first year (2006--2007) (winter and summer) showing that the reproductive period was restricted to few months per year. Thus, sampling was done yearly in the following sampling occasions (summer). For Portugal, no information about the reproductive period of either species was available. *A. nodosum*'s receptacles were present all year around and thus sampling was performed 3 times per year in 2005 (winter, spring and summer). In 2006 sampling was done twice a year (winter and summer). Additional in-situ observations showed that gamete release occurred mainly between October and February, thus between 2007 and 2009 sampling was done once a year (summer). For *F. serratus* in Portugal, sampling was done twice a year between 2006 and 2008 (winter and summer) and once a year in 2009, since mature receptacles were mainly observed between May and September (end of summer).
For the demographic analyses, individuals were divided into six classes: the first class included only the new recruits; individuals in this class can only grow to a larger class or die. The other classes were based on individual size (\<5 g; 5--15 g; 15--54 g; 54--190 g; \>190 g), following [@pone.0092177-berg2] and where individuals larger than 5 g were assumed to be fertile. It was not possible to associate recruits to a given fertile alga in the field. For that reason an indirect method was used to calculate the fertility values for a given size class. The individual fertility values were calculated as contributions to the observed number of recruits, for algae in size-classes 3--6 in proportion to their biomass. The assumption that fertility is proportional to adult biomass is supported by results in earlier papers [@pone.0092177-Svensson2], [@pone.0092177-berg3], [@pone.0092177-berg4], [@pone.0092177-berg5]. Projection matrices were constructed based on the life cycle graph shown in [Fig. 1](#pone-0092177-g001){ref-type="fig"}.
{#pone-0092177-g001}
Model {#s2c}
-----
### The matrix model {#s2c1}
Matrix population models are structured models where the population is divided into discrete stages, e.g. age, size or natural life-cycle stages or a combination of these types of stages. The population matrix describes the transition probability from one stage to another and the probability to stay in the same stage as well as the stage specific fertility. The change in population size can be projected in discrete time steps and there are numerous analyses that can be applied to the model (see [@pone.0092177-Caswell1] for a review). A stochastic matrix model is usually based on a finite set of projection matrices {**A** ~0~, **A** ~1~, ..., **A** ~k~}, each representing a specific environment (e.g. here year and country). The stochastic process can be described by a first-order Markov chain of length *T* (*t* = 0, 1, ..., *T*), where *T* is a large number (usually thousands of time steps). To model the population growth a matrix **A** ~t~ is drawn from the subset of matrices at each time *t*, with a set probability and multiplied by a column vector **n**(*t*) specifying the number of individuals in each stage class at time t. The basic stochastic matrix population model can be expressed as
where **n**(*t*) is a population vector and **A** ~t~ is the population projection matrix generated by the stochastic environmental process at time step *t*. If the probability distribution of the different environments does not change through time the environment is considered homogenous. The stochastic population growth rate can then be calculated as
To our knowledge there are no statistical methods to compare the population growth rate and thus is not possible to calculate the uncertainty in the estimation of this parameter.
### Model analyses and simulations {#s2c2}
Due to the unbalanced temporal sampling there are three matrices describing the first year and two for the second year for *A. nodosum* in Portugal. In year three and four there is only one matrix per year similarly to *A. nodosum* in France where there is only one matrix for each of the three years. For *F. serratus* in Portugal there are two matrices per year for the first and second year and one matrix for the third year. For *F. serratus* in France there are two matrices for the first year and one matrix per year for years two and three. For years with more than one matrix a periodic model was applied to calculate a yearly matrix [@pone.0092177-Caswell2]. The stochastic population growth rate was calculated using yearly matrices where all years had the same probability to appear in the stochastic sample path. For each species and site a mean matrix was calculated from the yearly matrices ([Table 1](#pone-0092177-t001){ref-type="table"}) e.g. for a site with three yearly matrices the mean matrix A~m~ = (A~1~+A~2~+A~3~)/3. The stable stage distribution and reproductive values were calculated as the right and left eigenvectors corresponding to the dominant eigenvalue of the mean matrix for each species and country.
10.1371/journal.pone.0092177.t001
###### Mean matrices from stochastic simulations of the *A. nodosum* and *F. serratus* populations in France and Portugal.
{#pone-0092177-t001-1}
--------------------------------------- -------- --------- -------- --------- -------- --------- -------- --------- --------- --------- ---------
***Ascophyllumnodosum*** **Portugal**
0 *F~12~* 0.0176 *F~13~* 0.2425 *F~14~* 0.6670 *F~15~* 1.7894 *F~16~* 5.2537
*G~21~* 0.3283 *L~22~* 0.5546 *S~23~* 0.1860 *S~24~* 0.1172 *S~25~* 0.2378 *S~26~* 0.5534
*G~31~* 0.0285 *G~32~* 0.1279 *L~33~* 0.4621 *S~34~* 0.2148 *S~35~* 0.1106 *S~36~* 0.0402
*G~41~* 0.0035 *G~42~* 0.0243 *G~43~* 0.2564 *L~44~* 0.5476 *S~45~* 0.3467 *S~46~* 0.1485
*G~51~* 0 *G~52~* 0.0008 *G~53~* 0.0140 *G~54~* 0.1108 *L~55~* 0.4672 *S~56~* 0.3278
*G~61~* 0 *G~62~* 0 *G~63~* 0.0002 *G~64~* 0.0012 *G~65~* 0.0172 *L~66~* 0.3603
***Ascophyllumnodosum*** **France**
0 *F~12~* 0 *F~13~* 0.0418 *F~14~* 0.1520 *F~15~* 0.4980 *F~16~* 2.2230
*G~21~* 0.2609 *L~22~* 0.3942 *S~23~* 0.1047 *S~24~* 0.0233 *S~25~* 0.0272 *S~26~* 0.0186
*G~31~* 0 *G~32~* 0.1049 *L~33~* 0.3809 *S~34~* 0.0853 *S~35~* 0.0615 *S~36~* 0.0344
*G~41~* 0 *G~42~* 0.0219 *G~43~* 0.3149 *L~44~* 0.4496 *S~45~* 0.1576 *S~46~* 0.0503
*G~51~* 0 *G~52~* 0 *G~53~* 0.0438 *G~54~* 0.2783 *L~55~* 0.4445 *S~56~* 0.2146
*G~61~* 0 *G~62~* 0 *G~63~* 0.0046 *G~64~* 0.0273 *G~65~* 0.1706 *L~66~* 0.4935
***Fucusserratus*** **Portugal**
0 *F~12~* 0.3104 *F~13~* 2.6177 *F~14~* 5.2028 *F~15~* 13.3929 *F~16~* 19.4457
*G~21~* 0.1458 *L~22~* 0.0825 *S~23~* 0.2734 *S~24~* 0.6538 *S~25~* 1.6188 *S~26~* 4.4581
*G~31~* 0.0063 *G~32~* 0.0312 *L~33~* 0.0456 *S~34~* 0.0411 *S~35~* 0.0400 *S~36~* 0.0102
*G~41~* 0.0022 *G~42~* 0.0368 *G~43~* 0.0849 *L~44~* 0.1273 *S~45~* 0.1064 *S~46~* 0.0749
*G~51~* 0.0008 *G~52~* 0.0220 *G~53~* 0.0804 *G~54~* 0.1615 *L~55~* 0.1295 *S~56~* 0.1266
*G~61~* 0.0004 *G~62~* 0.0015 *G~63~* 0.0111 *G~64~* 0.0367 *G~65~* 0.0955 *L~66~* 0.1136
***Fucusserratus*** **France**
0 *F~12~* 0.0306 *F~13~* 0.656 *F~14~* 1.886 *F~15~* 5.4206 *F~16~* 18.1057
*G~21~* 0.2305 *L~22~* 0.1496 *S~23~* 0.1692 *S~24~* 0.3163 *S~25~* 1.0242 *S~26~* 3.3693
*G~31~* 0.0608 *G~32~* 0.0905 *L~33~* 0.1667 *S~34~* 0.0729 *S~35~* 0.0457 *S~36~* 0.0092
*G~41~* 0.0278 *G~42~* 0.0614 *G~43~* 0.1825 *L~44~* 0.2797 *S~45~* 0.1488 *S~46~* 0.123
*G~51~* 0.0024 *G~52~* 0.0179 *G~53~* 0.1161 *G~54~* 0.2144 *L~55~* 0.2623 *S~56~* 0.305
*G~61~* 0 *G~62~* 0.0016 *G~63~* 0.0022 *G~64~* 0.0429 *G~65~* 0.1318 *L~66~* 0.2078
--------------------------------------- -------- --------- -------- --------- -------- --------- -------- --------- --------- --------- ---------
The transition codes are represented in all matrices to make easier the interpretation of the table. F: Fertility; G: Growth to larger size classes; S: Shrinkage; L: Loop.
### Sensitivity and elasticity {#s2c3}
The sensitivity and elasticity (proportional sensitivity) of the stochastic population growth rate to changes in the matrix elements were calculated using model simulation. Due to the unbalanced temporal sampling the methods in [@pone.0092177-Caswell3] and [@pone.0092177-berg6] could not be used directly but the simulations are equivalent to these methods. To calculate the sensitivity of the stochastic population growth rate a small figure (0.0001) was added to one element at the time in all the basic matrices. For years with more than one matrix, the yearly matrix was then calculated with the periodic model and added to the set of yearly matrices. A new stochastic population growth rate was calculated for the perturbed set of matrices using the same stochastic sample path as when calculating the original growth rate. This new stochastic population growth rate minus the original one divided by the size of the perturbation -1 gave the sensitivity for the matrix element that was perturbed. The matrices were then set back to their original values and the next element was perturbed until values for all elements were obtained. The size of the perturbation was decreased until the sensitivity values converged to three significant digits. The elasticity of the population growth rate was calculated in a similar way but with a factor close to unity (1.0001) instead of an absolute perturbation. All calculations and simulations were performed in MATLAB R2011a.
Results {#s3}
=======
Stochastic Population Growth Rate and Stable Stage Distribution {#s3a}
---------------------------------------------------------------
Model simulations showed that the stochastic population growth rate (λ*~s~*) varied between species and locations. When λ*~s~* is \>1 the average population size is growing and when it is \<1 the average population size is decreasing. λ*~s~* was higher for *A. nodosum* in Portugal (0.97) than in France (0.87) but lower for *F. serratus* in Portugal (0.77) than in France (1.05) ([Fig. 2](#pone-0092177-g002){ref-type="fig"}). All the λ*~s~* were slightly below 1, except for *F. serratus* in France ([Fig. 2](#pone-0092177-g002){ref-type="fig"}). For *A. nodosum*, variation in population growth rate was larger in France than in Portugal but the opposite trend was recorded for *F. serratus* ([Fig. 2](#pone-0092177-g002){ref-type="fig"}).
{#pone-0092177-g002}
The stable stage distribution calculated from the mean matrix showed for both *A. nodosum* and *F. serratus*, that populations in France had higher proportion of individuals belonging to the largest size classes (classes 5 and 6 for *A. nodosum* and classes 3 to 6 for *F. serratus*) than in Portugal. However, all the populations, especially *F. serratus*, both in Portugal and France, were dominated by smaller individuals belonging to classes 1 and 2 ([Fig. 3](#pone-0092177-g003){ref-type="fig"}).
{#pone-0092177-g003}
Reproductive Values {#s3b}
-------------------
Reproductive values increased with plant size for both species at both locations but more pronouncedly for *F. serratus*, especially in Portugal ([Fig. 4](#pone-0092177-g004){ref-type="fig"}). These results were related to the fertilities of each class and the transition probabilities between size classes. Higher fertilities were found in Portugal for both species at all size classes. However, for *A. nodosum* in Portugal the probability of transition to the class with higher fertility (class 6) is very low and the shrinkage to smaller size classes much higher than in France ([Table 1](#pone-0092177-t001){ref-type="table"}). On the contrary, for *F. serratus* the differences in the transition to higher size classes and in the shrinkage of individuals to smaller sizes between locations are not so pronounced, and do not balance the much higher values of fertility found in Portugal ([Table 1](#pone-0092177-t001){ref-type="table"}).
{#pone-0092177-g004}
Elasticity and Sensitivity {#s3c}
--------------------------
The pattern of elasticities of the stochastic population growth rate to changes in the matrix elements showed that the loop transition was the most important for the λ*~s~* of *A. nodosum* while the growth transition was the most important for the λ*~s~* of *F. serratus* ([Fig. 5](#pone-0092177-g005){ref-type="fig"}). Elasticities were very similar for *F. serratus* populations in France and in Portugal, with the shrinkage transition as the least important for λ*~s~* ([Fig. 5](#pone-0092177-g005){ref-type="fig"}). Fertility transitions were very low for *A. nodosum* when compared to *F. serratus* but were slightly higher in Portugal than in France ([Fig. 5](#pone-0092177-g005){ref-type="fig"}). For *A. nodosum*, loop transitions were more important in Portugal than in France while growth and shrinkage transitions showed the opposite geographical pattern ([Fig. 5](#pone-0092177-g005){ref-type="fig"}). The comparison between the elasticity across size classes, without taking into account the fertility, showed that for *A. nodosum* in France there was an important contribution of the 3 largest size classes (4--6) to λ*~s~* while in Portugal it was classes 2 to 4 that influence λ*~s~* the most ([Fig. 6](#pone-0092177-g006){ref-type="fig"}). For *F. serratus* the contribution of each size class to λ*~s~* was more evenly distributed among size classes 1 to 5 for both locations ([Fig. 6](#pone-0092177-g006){ref-type="fig"}).
{#pone-0092177-g005}
{#pone-0092177-g006}
The pattern of sensitivities of the stochastic population growth rate to changes in the matrix elements showed that, globally, there was a selection for individual growth in both species at both locations. However, for *A. nodosum* the sensitivity to changes in loop and shrinkage transitions were also important for population growth, while the sensitivity to changes in growth transition was totally dominant over the other matrix elements in *F. serratus* ([Fig. 7](#pone-0092177-g007){ref-type="fig"}). In general, the sensitivity of the stochastic population growth rate to changes in fertility was very low, in particular for *F. serratus*. For *A. nodosum*, fertility transition has higher sensitivity in Portugal than in France ([Fig. 7](#pone-0092177-g007){ref-type="fig"}). For *A. nodosum* the sensitivity to changes in growth and loop transitions were higher in Portugal while the shrinkage transition was slightly more important in France. The comparison between the sensitivities, without taking into account the fertility across size classes, showed that the sensitivity of the stochastic population growth rate was higher for changes in classes 2 to 4 for *A. nodosum* in Portugal, while in France it was in general lower and balanced between classes 2 and 6 ([Fig. 8](#pone-0092177-g008){ref-type="fig"}). For *F. serratus* sensitivity of λ*~s~* was much higher for changes in smaller classes (1 and 2) and higher in Portugal than in France ([Fig. 8](#pone-0092177-g008){ref-type="fig"}).
{#pone-0092177-g007}
{#pone-0092177-g008}
Discussion {#s4}
==========
This study revealed distinct long-term demographic trends for central and edge populations of *A. nodosum* and *F. serratus*, demonstrating the role of species identity and related life-history traits on the dynamics of edge populations. These results have wide implications for the understanding and predicting of biodiversity changes at range edges, by revealing that marginal versus central populations can have important demographic differences that determine their capacity for persistence or likelihood to shift their distributional limits.
*A. nodosum* and *F. serratus* are contrasting species in terms of life histories, habitats and genetic background. Earlier studies show that phenotypic traits are differentiated between *A. nodosum*'s southern-edge and central populations while *F. serratus* has similar traits at both locations but occupies a narrower vertical distribution in the south [@pone.0092177-Arajo1], [@pone.0092177-Pearson1]. The results of the present study show how these differences in fitness related traits are translated into the demography of the species and how sensitive the stochastic population growth rate is to changes in vital rates. An extension of this could be to use these models to predict population persistence under the future scenario of environmental variability posed by climate change.
Stochastic Population Growth Rate and Stable Stage Distribution {#s4a}
---------------------------------------------------------------
For *A. nodosum*, modeled λ*~s~* was lower and more variable at the center than at the edge of distribution, but below 1 for both locations. For *F. serratus* λ*~s~* was much lower and variable at southern than at central locations. In general, macroalgae populations have lower and more variable λ*~s~* in environmentally variable habitats as a consequence of high mortality of individuals [@pone.0092177-berg4], [@pone.0092177-Engelen1]. An increase in the mortality of *A. nodosum* individuals has been related to different sources of disturbance such as ice scouring, harvesting and trampling [@pone.0092177-berg4], [@pone.0092177-Ang1], [@pone.0092177-Arajo3]. The lower λ*~s~* found for the *A. nodosum* population in France was mainly the result of the lower survival of individuals in all size classes when compared to Portugal (see [Table 1](#pone-0092177-t001){ref-type="table"}). During this study, deposition of sand was frequently observed in some of the study plots in France but never in Portugal. This episodic disturbance might have accounted for the lower survival rates observed in France. A previous study in the same area of France in the late 1990′s found much lower values of mortality and never reported accumulation of sand in the study plots, but still found λ*~s~* of less than 1 due to low recruitment rates (Åberg *et al.* personal observation). Values of λ*~s~* below 1 were reported for *A. nodosum* in the northern part of the Swedish west coast, close to its distributional limit further south on the west coast [@pone.0092177-Svensson2]. For central populations, λ*~s~* varied from slightly below to slightly above 1, depending on the location and sampling dates [@pone.0092177-Svensson2], (Åberg et al. personal observation). This demonstrates the need to conduct long-term observational studies covering a comprehensive spatial extent, along the distributional range of species, when constructing long-term predictions about population's trends.
Southern range *F. serratus*'*s* populations show reduced plasticity to environmental variation [@pone.0092177-Pearson1]. In the present study, model simulations showed, for *F. serratus*, that environmental fluctuation at the boundary of its distribution, might cause its disappearance from previously colonized areas, which take a long time to recover [@pone.0092177-Arrontes2]. The λ*~s~* values found in this study might reflect the maladaptation of populations to years with high environmental variation.
For both species, large individuals were much more abundant in France than in Portugal. The individual growth rate of both species is similar at both locations (Araújo et al. personal observation), explaining that the lower transition probability to larger size classes in Portugal is caused by higher breakage of fronds. In previous studies, increased breakage of macroalgae's fronds has been related with several factors such as grazing [@pone.0092177-Toth1], [@pone.0092177-Viejo2], [@pone.0092177-Davies1], [@pone.0092177-Toth2], wave action [@pone.0092177-Engkvist1], [@pone.0092177-Kraufvelin1], harvesting [@pone.0092177-AngJr1] and trampling [@pone.0092177-Arajo3]. However, the comparative quantitative effects of most of these factors on breakage of the fronds at central and southern locations are unknown for these species.
Reproductive Values {#s4b}
-------------------
The reproductive values increased with individual size for both species but more pronouncedly for *F. serratus*, especially in Portugal. This indicates that *F. serratus* individuals in Portugal have a higher fitness advantage in terms of reproduction and survival if they grow to larger size classes. An increase in the reproductive biomass with size of the plant is a common trend for seaweeds [@pone.0092177-berg3], [@pone.0092177-AngJr2] and an increase in the reproductive value with size class has been reported in previous studies [@pone.0092177-berg2], [@pone.0092177-Engelen1], [@pone.0092177-AngJr1]. Reproductive effort might increase following disturbance [@pone.0092177-Marba1], [@pone.0092177-Welham1], [@pone.0092177-Alexandre1]. The high reproductive values of *F. serratus* in this study might reflect the levels of environmental disturbance found at its edge of distribution.
Elasticity {#s4c}
----------
Elasticity of the population growth rate to changes in survival was higher than for changes in fertility, for both species at both locations. *A. nodosum* was more susceptible to changes in loop transition followed by growth in both areas, while changes in growth transitions were the most important for *F. serratus* followed by loop. However, *F. serratus* populations were more susceptible to changes in fertility rates than those of *A. nodosum*. In general for algae, higher elasticity values for survival than for fertility are reported for longer-lived species, while in shorter-lived species elasticity values for fertility, although lower than for survival, are generally higher [@pone.0092177-Svensson2], [@pone.0092177-berg4], [@pone.0092177-Ang1], [@pone.0092177-Engel1], [@pone.0092177-AngJr3].
The higher susceptibility to changes in loop transitions of *A. nodosum* populations demonstrates that their long-term persistence might be affected by factors increasing the shrinkage of individuals (in particular of adults). Among these factors, an increase in the human frequentation of coastal areas might be important since human trampling negatively affects the communities at the southern-limit of *A. nodosum,* both in the short and long term [@pone.0092177-Arajo3]. The density of invertebrates inhabiting *A. nodosum* adults is lower in Portugal than in central and northern populations [@pone.0092177-Arajo1], [@pone.0092177-Viejo2]. However, in spite of the decreasing trend in limpet grazing from north to south [@pone.0092177-Coleman1], the opposite trend was recorded under the *A. nodosum* canopy when comparing edge and central populations [@pone.0092177-Arajo1]. Harvesting, although not important in Portugal, occurs in France, and increases individual shrinkage [@pone.0092177-Ang1], [@pone.0092177-Ang2]. Also changes in the frequency of storms predicted by models of climate change [@pone.0092177-Allen1] or in the pressure of grazing [@pone.0092177-Hawkins1] might account for increased shrinkage of individuals. Although the *A. nodosum* populations in Portugal were slightly more susceptible to changes in fertility than in France, elasticities for fertility transitions were low. The reproductive output [@pone.0092177-Arajo1] and reproductive allocation (Araújo et al. personal observation) of *A. nodosum* was higher in Portugal than in France, despite the apparent low importance for population growth and persistence demonstrated by the model. However, in long-lived algae, recruitment is important for the long-term stability of populations [@pone.0092177-AngJr1], [@pone.0092177-Reed1], and massive recruitment may occur under exceptional circumstances such as disturbance events [@pone.0092177-Engelen1], [@pone.0092177-Ladah1]. Short-term studies, like the one presented here, are likely to miss these extraordinary recruitment events, potentially contributing to the occurrence of estimates of population growth below 1 observed in this study.
Factors affecting growth of individuals also have negative effects on the growth of the population, as demonstrated by the high elasticities of growth transition found for both species. The documented North Atlantic trend of raising sea and air temperatures [@pone.0092177-Easterling1] is expected to have more immediate and severe consequences at species borders. As expected, temperatures in the intertidal area are higher in Portugal than in France [@pone.0092177-Arajo1]. Increased sea surface temperature will decrease *A. nodosum* growth at temperatures above 19°C with lethal effects shown above 28°C [@pone.0092177-Keser1]. *F. serratus* southern edge populations show much lower resilience to heat stress than central ones [@pone.0092177-Pearson1]. Consequently, it can be expected that, under the current scenario of global warming, these southern limit populations will suffer a decrease in individual growth that could have important consequences for population persistence.
Sensitivity {#s4d}
-----------
The sensitivity of the stochastic population growth rate to changes in a given vital rate can be interpreted as the selection pressure on that life-history transition [@pone.0092177-Caswell1]. For both species and for both locations, the high sensitivity values of the population growth rate to changes in transition to larger size classes indicate that there is strong selection for individual growth. This selective pressure on individual growth is very pronounced for *F. serratus*, dominating over all the other transitions. For short-lived species, increased growth, directly related for *F. serratus* with its high reproductive values in large size classes, might maximize the individual life time fertility. In a scenario of increased environmental and anthropogenic disturbance, with negative effects on individual plant growth, the selective pressure for increased individual growth will be stronger. When under stressful conditions, the most beneficial trait for population persistence, in this case individual growth, will be favoured at the expenses of other fitness related traits like reproduction and defence [@pone.0092177-Obeso1], [@pone.0092177-HairstonJr1]. However, a reduction in the allocation of resources to defence will increase individual growth rates [@pone.0092177-Pavia1] whilst also increasing the risk of both breakage and grazing, and thus decreased net growth [@pone.0092177-Pavia1], [@pone.0092177-Toth3]. The selective pressure on fertility is higher in Portugal than in France, although much lower than for survival transitions. Nevertheless, if the allocation of resources for growth increases, the reproductive output might be reduced with consequences for the long-term population persistence.
The results of this study show that *A. nodosum* has high plasticity in life-history traits, while *F. serratus*, with much lower and more variable λ*~s~*, seems to be more susceptible to environmental variation in local habitat conditions at the southern limit boundary. For both species, changes affecting individual growth (by diminishing the transition to larger size classes or increasing shrinkage) would reduce population growth. Under the current scenario of global environmental change, anthropogenic and environmental disturbances affecting vital rates may condition the local persistence of populations in the long term.
Future research should include studies with a particular focus on the southern biogeographic range limit of species distribution, addressing the comparison of phenotypic fitness related traits in geographically isolated versus connected edge populations.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: RA ES IS PA. Performed the experiments: RA. Analyzed the data: RA PA. Contributed reagents/materials/analysis tools: ES IS. Wrote the paper: RA ES IS PA.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Accurate identification of the translation initiation codon in mRNA by ribosomes is crucial for expression of the correct cellular proteins. This process generally occurs in eukaryotic cells by a scanning mechanism, wherein the small (40S) ribosomal subunit recruits charged initiator tRNA (Met-tRNA~i~^Met^) in a ternary complex (TC) with eIF2-GTP, and the resulting 43S pre-initiation complex (PIC) attaches to the 5' end of the mRNA and scans the 5'UTR for an AUG start codon. In the scanning PIC, the TC is bound in a relatively unstable state, dubbed 'P~OUT~', suitable for inspecting successive triplets in the P decoding site for perfect complementarity with the anticodon of Met-tRNA~i.~ Nucleotides surrounding the AUG, particularly at the −3 and +4 positions (relative to the AUG at +1 to +3), the 'Kozak context', also influence the efficiency of start codon recognition. Hydrolysis of the GTP bound to eIF2 can occur, dependent on GTPase activating protein eIF5, but P~i~ release is blocked by eIF1, whose presence also prevents highly stable binding of Met-tRNA~i~^Met^ in the 'P~IN~' state. Start-codon recognition triggers dissociation of eIF1 from the 40S subunit, which in concert with other events allows P~i~ release from eIF2-GDP·P~i~ and accommodation of Met-tRNA~i~^Met^ in the P~IN~ state of the 48S PIC ([Figure 1A](#fig1){ref-type="fig"}). Subsequent dissociation of eIF2-GDP and other eIFs from the 48S PIC enables eIF5B-catalyzed subunit joining and formation of an 80S initiation complex with Met-tRNA~i~^Met^ base-paired to AUG in the P site (reviewed in \[[@bib23]\] and \[[@bib24]\]). eIF1 plays a dual role in the scanning mechanism, promoting rapid TC loading in the P~OUT~ conformation while blocking rearrangement to P~IN~ at non-AUG codons by clashing with Met-tRNA~i~ in the P~IN~ state ([@bib55]; [@bib37])([@bib26]), thus requiring dissociation of eIF1 from the 40S subunit for start codon recognition ([Figure 1A](#fig1){ref-type="fig"}). Consequently, mutations that weaken eIF1 binding to the 40S subunit reduce the rate of TC loading and elevate initiation at near-cognate codons (eg. UUG), or AUG codons in poor context, by destabilizing the open/P~OUT~ conformation and favoring rearrangement to the closed/P~IN~ state during scanning ([@bib43]; [@bib44]). Moreover, decreasing wild-type (WT) eIF1 abundance reduces initiation accuracy, whereas overexpressing eIF1 suppresses initiation at near-cognates or AUGs in poor context ([@bib63]; [@bib3]; [@bib28]; [@bib57]; [@bib43]). The mechanistic link between eIF1 abundance and initiation accuracy is exploited to negatively autoregulate eIF1 expression, as the AUG codon of the eIF1 gene (*SUI1* in yeast) occurs in suboptimal context and the frequency of its recognition is inversely related to eIF1 abundance ([@bib28]; [@bib43]). Mutations that weaken 40S binding by eIF1 relax discrimination against the poor context of the *SUI1* AUG codon and elevate eIF1 expression, overcoming autoregulation ([@bib43]). In contrast, mutations that enhance eIF1 binding to the 40S subunit impede rearrangement of the scanning PIC to the closed/P~IN~ conformation ([@bib43]; [@bib45]), which increases discrimination against the poor context of the *SUI1* AUG codon, to reduce eIF1 expression, and also suppresses initiation at near-cognate UUG codons ([@bib43]; [@bib45]).
![eIF1A-NTT residues associated with UM are predicted to participate in stabilizing the P~IN~ state of the closed conformation of the 48S PIC.\
(**A**) Model describing known conformational rearrangements of the PIC during scanning and start codon recognition. (i) eIF1 and the scanning enhancers (blue balls) in the C-terminal tail (CTT) of eIF1A stabilize an open conformation of the 40S subunit to which TC rapidly binds. (ii) The 43S PIC in the open conformation scans the mRNA for the start codon with Met-tRNA~i~^Met^ bound in the P~OUT~ state. eIF2 can hydrolyze GTP to GDP•P~i~, but release of P~i~ is blocked by eIF1. The N-terminal tail (NTT) of eIF1A interacts with the eIF5-CTD. (iii) On AUG recognition, Met-tRNA~i~^Met^ moves from the P~OUT~ to P~IN~ state, clashing with eIF1 and the CTT of eIF1A, provoking displacement of the eIF1A CTT from the P site, dissociation of eIF1 from the 40S subunit, and P~i~ release from eIF2. The NTT of eIF2β interacts with the eIF5-CTD, and the eIF1A-NTT, harboring scanning inhibitor (SI) elements, interacts with the codon:anticodon helix. (Above) Arrows summarize that eIF1 and the eIF1A SE elements promote P~OUT~ and impede transition to P~IN~ state, whereas the eIF1A SI element in the NTT stabilizes the P~IN~ state. (Adapted from ([@bib23])). Results presented below show that this function of the eIF1A-NTT is impaired by uveal melanoma (UM)-associated substitutions and others that disrupt direct contacts with the mRNA or codon:anticodon helix shown in (**B**). (**B**) Magnified portion of the py48S PIC structure (PDB 3J81) showing contacts made by the eIF1A-NTT (shades of blue and cyan) in the closed/P~IN~ conformation. Side-chains of NTT residues substituted in UM (red) or directly contacting 18S rRNA (yellow), tRNA~i~ (green) or mRNA (purple) are shown as sticks. (**C**) Sequence of human eIF1A NTT residues 1--15 showing the substitutions (red) or deletions (dash) found in the indicated UM tumors. Substitutions in yeast eIF1A corresponding to those found in UM tumors are listed on the last line. The five basic residues of the yeast NTT making direct contacts in the PIC and substituted here in addition to the UM-associated substitutions are shown in cyan.](elife-31250-fig1){#fig1}
eIF1A also has a dual role in scanning and start codon recognition. Scanning enhancer (SE) elements in the eIF1A C-terminal tail (CTT) promote TC binding in the open P~OUT~ conformation and impede rearrangement to the closed P~IN~ state, such that substitutions that impair the SE elements both impair TC recruitment and increase initiation at near-cognate start codons ([@bib57]). Biochemical mapping experiments suggest that, like eIF1, the eIF1A CTT clashes with Met-tRNA~i~ in the P~IN~ state ([@bib67]), and is displaced from the P site on start codon recognition ([@bib68]) to enable a functional interaction of the eIF1A CTT with the NTD of eIF5, the GTPase activating protein for eIF2, that facilitates P~i~ release from eIF2-GDP·P~i~ ([@bib50]). Scanning inhibitor elements SI~1~ and SI~2~ in the unstructured eIF1A N-terminal tail (NTT) and helical domain, respectively, antagonize SE function and stabilize the closed/P~IN~ conformation on start codon recognition ([Figure 1A](#fig1){ref-type="fig"}). Accordingly, substitutions that impair SI elements destabilize the closed complex and accelerate TC loading to the open complex in vitro, and promote continued scanning at UUG codons in hypoaccurate mutant cells ([@bib15]) ([@bib57]). SI~1~ mutations also increase the probability that the scanning PIC will bypass an upstream AUG codon (leaky scanning) ([@bib15]; [@bib39]); and one such mutation, substituting NTT residues 17--21, decreases recognition of the suboptimal AUG codon of *SUI1* mRNA to reduce eIF1 expression ([@bib43]).
Molecular insight into the deduced function of the eIF1A-NTT of promoting AUG recognition during scanning came from the cryo-EM structure of a partial yeast 48S PIC (py48S) containing eIF1, eIF1A, TC and mRNA, with the Met-tRNA~i~ base-paired to the AUG codon in a P~IN~ state. All but the first four residues of the eIF1A NTT were visible in this structure, and basic NTT residues Lys7, Lys10, Arg13, and Lys16 contact either the anticodon or the +4 to +6 mRNA nucleotides adjacent to the AUG codon, while Arg14 interacts with the 18S rRNA ([Figure 1B](#fig1){ref-type="fig"}) ([@bib26]). These findings suggest that the eIF1A NTT can directly stabilize the P~IN~ state, and help to explain how NTT substitutions in SI~1,~ which spans residues 1--26 ([@bib57]), increase discrimination against non-AUG codons, which form less stable codon:anticodon duplexes than do AUG codons. Other studies have implicated NTT residues 7--11 and 12--16, encompassing the aforementioned basic NTT residues, in interactions with eIF1 and the eIF5-CTD that appear to promote assembly of the open, scanning PIC ([@bib15]; [@bib39]). The β-subunit of eIF2 also harbors a highly basic NTT, which interacts with the eIF5-CTD to promote eIF1 dissociation from the closed complex at the start codon ([@bib38]). It was suggested that interaction of the eIF5-CTD with the eIF1A-NTT would stabilize the open conformation of the PIC prior to AUG recognition, whereas alternative interaction of the eIF5-CTD with the eIF2β-NTT would stabilize the closed conformation of the PIC on AUG recognition ([@bib39]). The proposed dissociation of the eIF1A-NTT from the eIF5-CTD on AUG recognition would free the eIF1A-NTT for interactions with the mRNA and anticodon evident in the py48S PIC ([@bib26]). Thus, the eIF1A-NTT would play a dual role of promoting the open conformation of the PIC through interaction with the eIF5-CTD and subsequently stabilizing the closed state by interacting with the mRNA and anticodon.
Somatic mutations in the human gene *EIF1AX* encoding eIF1A are frequently associated with uveal melanomas (UM) associated with disomy for chromosome 3, and all of the *EIF1AX* mutations sequenced thus far produce in-frame substitutions or deletions of one or more residues in the first 15 residues of the NTT ([@bib42]). A subset of these mutations substitute or delete two of the four basic residues that contact mRNA or the tRNA~i~ anticodon in the yeast py48S PIC (Lys7 and Arg13), others introduce acidic residues that might electrostatically repel the phosphodiester backbone of the mRNA or tRNA~i~, while others affect a Gly-Gly turn that is important for correct positioning of the basic residues in the PIC ([Figure 1B--C](#fig1){ref-type="fig"}) ([@bib26]). Thus, all of the UM mutations might affect eIF1A function by the same mechanism, of weakening the ability of the eIF1A NTT to stabilize the P~IN~ conformation of the tRNA~i~. As such, they would be expected to increase discrimination against poor initiation sites in vivo. Alternatively, they could impair the proposed function of the eIF1A-NTT in stabilizing the open conformation ([@bib39]), in which case they would have the opposite effect and relax discrimination against suboptimal start codons. We set out to distinguish between these possibilities by examining the consequences of seven yeast eIF1A-NTT substitutions equivalent to those associated with UM in residues Lys3, Lys4, Thr6, Gly8, Arg13 and Gly15, and also of altering the five NTT basic residues that interact with the mRNA or anticodon in the py48S PIC (Lys7, Lys10, Arg13, Arg14 and Lys16) ([Figure 1C](#fig1){ref-type="fig"}). Our genetic and biochemical analyses indicate that UM-associated eIF1A substitutions disrupt NTT interactions with the mRNA or tRNA~i~ to destabilize the closed/P~IN~ conformation of the PIC and increase discrimination against near-cognate codons or AUGs in suboptimal context, with particularly strong effects observed for substitutions of Arg13---one of five basic residues that interacts with the mRNA/tRNA~i~ anticodon. Ribosome profiling of the potent UM-associated mutant eIF1A-R13P reveals widespread increased discrimination against AUG codons in poor context, which can alter recognition of the start codon for the main coding sequences (CDS) or indirectly affect translation by modulating recognition of upstream open reading frames (uORFs) in the mRNA leader. These findings allow us to predict that eIF1A-NTT mutations alter gene expression in UM tumor cells by shifting translation initiation at main CDS and regulatory uORFs from poor to optimum initiation sites.
Results {#s2}
=======
UM-associated substitutions in the yeast eIF1A NTT increase discrimination against near-cognate UUG codons in vivo {#s2-1}
------------------------------------------------------------------------------------------------------------------
To explore functional consequences of substitutions in human eIF1A associated with uveal melanoma ([@bib42]), we introduced substitutions into the yeast eIF1A NTT corresponding to 7 of the 13 substitutions associated with the disease: K3D, K4D, T6R, T6D, ΔG8, R13P, and G15D ([Figure 1C](#fig1){ref-type="fig"}). Asn4 and Gly6 of human eIF1A correspond to Lys4 and Thr6 in yeast, thus the yeast K4D and T6R/T6D substitutions mimic the human N4D and G6R/G6D UM-associated substitutions, respectively. The deletion of Gly8 (ΔG8) in yeast produces the same protein as the UM-associated substitution ΔG9, leaving a single Gly residue in place of the Gly8/Gly9 pair ([Figure 1C](#fig1){ref-type="fig"}). Mutations were generated in a *TIF11* allele (encoding yeast eIF1A) under the native promoter and the mutant alleles on single-copy plasmids were used to replace WT *TIF11^+^* by plasmid-shuffling in a *his4-301* yeast strain in order to examine their effects on initiation at near-cognate UUG codons. *his4-301* lacks an AUG start codon and confers histidine auxotropy, which can be suppressed by mutations in initiation factors with the Sui^-^ phenotype ([Su]{.ul}ppressor of [i]{.ul}nitiation codon mutation) that increase initiation at the third, in-frame UUG codon to restore expression of histidine biosynthetic enzyme His4 ([@bib12]). None of the *TIF11* mutations allows detectable growth on medium containing only 1% of the usual histidine supplement ([Figure 2---figure supplement 1A](#fig2s1){ref-type="fig"}, -His medium), indicating the absence of Sui^-^ phenotypes; and none confers a slow-growth phenotype (Slg^-^) on complete medium ([Figure 2---figure supplement 1A](#fig2s1){ref-type="fig"}, +His). We next tested the mutant alleles for the ability to suppress the elevated UUG initiation on *his4-301* mRNA and the attendant His^+^ phenotype conferred by dominant Sui^-^ mutations *SUI5* and *SUI3-2* encoding, respectively, the G31R variant of eIF5 and S264Y variant of eIF2β ([@bib25]). Remarkably, the dominant His^+^ phenotypes conferred by plasmid-borne *SUI5* or *SUI3-2* are diminished by all of the NTT mutations ([Figure 2A](#fig2){ref-type="fig"} and [Figure 2---figure supplement 1B](#fig2s1){ref-type="fig"}, -His); and the Slg^-^ phenotype conferred by *SUI5* in +His medium at 37°C is also suppressed by the *K3E, K4D, ΔG8, R13P,* and *G15D* mutations ([Figure 2A](#fig2){ref-type="fig"}, +His, 37°C). These results suggest that the UM-associated substitutions, as a group, mitigate the effects of *SUI5* and *SUI3-2* in elevating UUG initiation, increasing discrimination against near-cognate start codons.
{ref-type="fig"} and [Figure 2---figure supplement 1](#fig2s1){ref-type="fig"}.Effects of UM-associated substitutions in the yeast eIF1A NTT on *HIS4-lacZ* UUG:AUG expression ratios in *SUI3-2* cells, eIF1 levels, expression of *SUI1-lacZ* fusions containing the native, poor or optimum AUG contexts and *GCN4-lacZ* expression in *SUI3-2* cells.](elife-31250-fig2){#fig2}
The effect of *SUI3-2* in reducing the fidelity of start codon selection can be quantified by measuring the expression of matched *HIS4-lacZ* reporters containing a UUG or AUG triplet as start codon. As expected ([@bib25]), *SUI3-2* increases the ratio of expression of the UUG to AUG reporter from the low WT value of \~3% up to \~12% ([Figure 2B](#fig2){ref-type="fig"}). With the exception of *T6D*, all of the UM mutations significantly reduced the *HIS4-lacZ* UUG:AUG initiation ratio, with *R13P* eliminating \~75% of the increase in the UUG/AUG initiation ratio conferred by *SUI3-2* in *TIF11^+^* cells ([Figure 2B](#fig2){ref-type="fig"}). The results indicate that eIF1A UM substitutions restore to varying extents discrimination against near-cognate UUG codons in Sui^-^ mutants, thus conferring Ssu^-^ phenotypes.
Many Sui^-^ mutations, including *SUI3-2,* derepress *GCN4* mRNA translation in nutrient-replete cells (the Gcd^-^ phenotype) ([@bib57]). This phenotype generally results from a reduced rate of TC recruitment that allows 40S subunits that have translated upstream open reading frame 1 (uORF1) and resumed scanning to subsequently bypass uORFs 2--4 and reinitiate at the *GCN4* AUG codon in the absence of a starvation-induced inhibition of TC assembly ([@bib21]). Interestingly, the Gcd^-^ phenotype of *SUI3-2,* manifested as an \~3 fold derepression of a *GCN4-lacZ* reporter, is also significantly diminished by *R13P* ([Figure 2---figure supplement 1C](#fig2s1){ref-type="fig"}), the eIF1A NTT mutation shown above to be the strongest suppressor of the Sui^-^ phenotype of *SUI3-2* ([Figure 2B](#fig2){ref-type="fig"}). Co-suppression of the Gcd^-^ and Sui^-^ phenotypes of *SUI3-2* has been demonstrated for other Ssu^-^ mutations in eIF1A ([@bib57]; [@bib13]; [@bib45]) and attributed to destabilization of the closed/P~IN~ conformation and attendant shift to the open scanning-conducive conformation to which TC binds rapidly ([@bib52]). Thus, co-suppression of the Gcd^-^ and Sui^-^/hypoaccuracy phenotypes of *SUI3-2* observed only for the *R13P* mutation suggests that it exceeds the other UM-associated mutations in destabilizing the closed/P~IN~ conformation of the PIC.
UM-associated eIF1A substitutions increase discrimination against AUG codons in poor context {#s2-2}
--------------------------------------------------------------------------------------------
In addition to reducing initiation at near-cognate UUG codons in Sui^-^ mutants, Ssu^-^ substitutions in eIF1 and eIF1A are known to increase discrimination against the AUG start codon of the *SUI1* gene encoding eIF1, which exhibits a non-preferred Kozak context. The unfavorable context of the *SUI1* start codon underlies negative autoregulation of eIF1 synthesis, which hinders overexpression of eIF1 as excess eIF1 impedes initiation at its own start codon ([@bib43]). Consistent with this, the eIF1A UM mutations reduce the steady-state level of eIF1, with the strongest reduction for *R13P,* lesser reductions for *K3E, K4D, ΔG8,* and *G15D,* and the smallest effects for *T6R* and *T6D* ([Figure 2C](#fig2){ref-type="fig"}, eIF1 blot and eIF1/Gcd6 ratios). This hierarchy exactly parallels that observed for suppression of the UUG:AUG initiation ratio in *SUI3-2* cells for these eIF1A mutants ([Figure 2B](#fig2){ref-type="fig"}).
Results in [Figure 2C](#fig2){ref-type="fig"} also reveal that *K4D, ΔG8, T6R* and *T6D* reduce expression of eIF1A itself (eIF1A blot). It seems unlikely that these reductions arise from altered translation of eIF1A, as the eIF1A AUG codon is in good context (A at −3) and the reductions do not correlate with decreases in eIF1 expression conferred by different eIF1A variants ([Figure 2C](#fig2){ref-type="fig"}). Rather, these substitutions, and those at Lys10 discussed below (Figure 4A), might impair a role of the first 10 residues of eIF1A in stabilizing the protein. Regardless, the reduced expression of these eIF1A variants is insufficient to confer a marked reduction in eIF1 synthesis or a strong Ssu^-^ phenotype, as both eIF1A-T6R and eIF1A-T6D are poorly expressed but have a small impact on both eIF1 expression ([Figure 2C](#fig2){ref-type="fig"}) and the enhanced UUG initiation conferred by *SUI3-2* ([Figure 2B](#fig2){ref-type="fig"}). Furthermore, we show below that increasing the expression of the eIF1A-K4D and -ΔG8 variants does not diminish their effects on UUG initiation or eIF1 expression.
In accordance with their effects on eIF1 expression, the *R13P, K3E, K4D, ΔG8,* and *G15D* mutations significantly reduce expression of the WT *SUI1-lacZ* fusion containing the native, poor context of the eIF1 AUG codon, ~-3~CGU~-1~ ([Figure 2D](#fig2){ref-type="fig"}, Native context). These eIF1A mutations also reduce expression of a second reporter in which the native AUG context is replaced with the even less favorable context of ~-3~UUU~-1~ ([@bib43]), with *R13P* again conferring the largest reduction ([Figure 2D](#fig2){ref-type="fig"}, poor context). Importantly, none of the mutations significantly affects expression of a third reporter containing the optimum context of ~-3~AAA~-1~ ([@bib43]) ([Figure 2D](#fig2){ref-type="fig"}). Thus, a subset of the UM mutations, and particularly *R13P* and *G15D,* selectively reduce recognition of the eIF1 AUG codon when it resides in its native poor context, or in another poor context, in addition to increasing discrimination against near-cognate UUG start codons.
NTT basic residues contacting mRNA or tRNA~i~ in the py48S complex have a role in recognition of poor initiation sites in vivo {#s2-3}
------------------------------------------------------------------------------------------------------------------------------
Among the UM mutations, *R13P* consistently conferred the greatest reduction in recognition of both UUG codons and AUGs in poor context ([Figure 2](#fig2){ref-type="fig"} and [Figure 2---figure supplement 1](#fig2s1){ref-type="fig"}). In the structure of py48S, Arg13 contacts the +5 nucleotide in mRNA, and with Lys7, Lys10, and Lys16, is one of four basic residues in the eIF1A NTT contacting the mRNA or tRNA~i~ anticodon ([Figure 1B](#fig1){ref-type="fig"}). A fifth basic residue, Arg14 contacts A1427/G1428 of 18S rRNA located in the mRNA binding cleft ([@bib26]). In addition, UM mutation *ΔG8* affects the tandem Gly8-Gly9 pair that mediates a turn in the NTT required for proper positioning of the four basic residues. Accordingly, we hypothesized that the hyperaccuracy phenotypes of the UM-associated substitutions R13P and ΔG8 reflect loss of a direct contact with the mRNA (R13P) or perturbation of one or more contacts of the four basic residues with mRNA/tRNA~i~ (*ΔG8*), which destabilizes the P~IN~ state of the 48S PIC. Moreover, insertion of an acidic side-chain between basic residues Arg14 and Lys16 by UM substitution G15D ([Figure 1B](#fig1){ref-type="fig"}), which could introduce electrostatic repulsion with the backbone of mRNA or rRNA, could likewise destabilize the 48S PIC. Because UUG start codons form a less stable codon:anticodon helix with a U:U mismatch compared to the perfect duplex formed at AUG codons, UM substitutions that destabilize P~IN~ should be especially deleterious to initiation at UUG codons, as we observed ([Figure 2](#fig2){ref-type="fig"}). To test this hypothesis, we introduced Ala and Asp substitutions at all five of the NTT basic residues that contact mRNA, tRNA~i~ or rRNA in the py48S PIC ([@bib26]), expecting to find stronger hyperaccuracy phenotypes for Asp versus Ala substitutions owing to electrostatic repulsion with the nucleic acids in the case of Asp replacements. We also examined a double deletion of Gly8-Gly9 that we reasoned might have a stronger phenotype than the UM mutation *ΔG8.*
We observed modest Slg^-^ phenotypes for the R13D and R14D substitutions, but no His^+^ phenotypes indicative of Sui^-^ defects for any of the targeted NTT mutations ([Figure 3---figure supplement 1A](#fig3s1){ref-type="fig"}). Remarkably, both Ala and Asp substitutions of Lys10, Arg13, Arg14, and Lys16, and the Asp substitution of Lys7, all diminished the His^+^/Sui^-^ phenotype of *SUI3-2* ([Figure 3A](#fig3){ref-type="fig"}) and decreased the *HIS4-lacZ* UUG:AUG initiation ratio in *SUI3-2* cells, with the greatest reductions seen for R13D, R14D, and K16D. In agreement with our hypothesis, the Asp versus Ala substitutions generally conferred greater suppression of the UUG:AUG ratio, but especially so at Lys10 and Lys16 ([Figure 3B](#fig3){ref-type="fig"}). Using a second set of UUG and AUG reporters, expressing renilla or firefly luciferase from different transcripts under the control of the *ADH1* (*RLUC*) or *GPD* (*FLUC*) promoter, we confirmed that the K16D and R13P substitutions reduced the elevated UUG:AUG initiation ratio conferred by *SUI3-2* ([Figure 3---figure supplement 1B](#fig3s1){ref-type="fig"}). All of the mutations, except for K7A, also diminished the Gcd^-^ phenotype of *SUI3-2*, reducing the derepression of *GCN4-lacZ* expression, again with generally greater reductions for Asp versus Ala replacements ([Figure 3C](#fig3){ref-type="fig"}). The degree of suppression of the elevated UUG:AUG ratio and *GCN4-lacZ* expression in *SUI3-2* cells was correlated, with *R13D, R14D,* and *K16D* being the strongest suppressors of both phenotypes (cf. [Figure 3B and C](#fig3){ref-type="fig"}). As noted above, this co-suppression of impaired TC loading (Gcd^-^) and increased UUG recognition (Sui^-^) phenotypes suggest that these eIF1A NTT substitutions specifically destabilize the closed/P~IN~ state with attendant shift to the open/P~OUT~ scanning conformation of the PIC ([@bib57]).
{ref-type="fig"}. (**B--C**) Transformants of the strains from (**A**) harboring *HIS4-lacZ* reporters with AUG or UUG start codons (**B**) or the *GCN4-lacZ* reporter (**C**) were cultured and assayed for β-galactosidase activities as in [Figure 2B](#fig2){ref-type="fig"}.\
10.7554/eLife.31250.008Figure 3---source data 1.Source data for [Figure 3](#fig3){ref-type="fig"} and [Figure 3---figure supplement 1](#fig3s1){ref-type="fig"}.Effects of substitutions in eIF1A NTT basic residues on *HIS4-lacZ* UUG:AUG expression ratios and *GCN4-lacZ* expression both in *SUI3-2* cells. Effects of eIF1A-R13P and K16D substitutions on *FLUC/RLUC* UUG:AUG expression ratios in *SUI3-2* cells.](elife-31250-fig3){#fig3}
In addition to suppressing UUG initiation, all of the targeted substitutions of the five basic residues, and the deletion of Gly8-Gly9, also increase discrimination against the non-preferred context of the eIF1 AUG codon, reducing expression of eIF1 ([Figure 4A](#fig4){ref-type="fig"}) and of the *SUI1-lacZ* fusions with native or poor context, without altering expression of *SUI1-lacZ* with optimal AUG context ([Figure 4B](#fig4){ref-type="fig"}). Again, the Asp versus Ala substitutions of the basic NTT residues generally confer stronger phenotypes ([Figure 4A--B](#fig4){ref-type="fig"}), consistent with stronger disruptions of NTT contacts with mRNA, tRNA~i~ or rRNA on introduction of negatively charged side-chains.
{ref-type="fig"}. (**B**) Transformants of strains in (**A**) with *SUI1-lacZ* fusions containing the native suboptimal, poor or optimum AUG contexts were assayed for β-galactosidase activities as in [Figure 2D](#fig2){ref-type="fig"}. (**C**) H3582 derivatives, harboring WT, *R13P* or *R14A TIF11* alleles and *el.uORF1 GCN4-lacZ* reporters containing the depicted optimum (pC3502, row1), weak (pC4466, row2) or poor (pC3503, row3) context of uAUG-1, or uORF-less *GCN4-lacZ* reporter with a mutated uAUG-1 (pC3505, row4), were assayed for β-galactosidase activities as in [Figure 2D](#fig2){ref-type="fig"}. Mean expression values with S.E.M.s were determined from four transformants (columns 1, 2 and 3). The percentages of scanning ribosomes that translate el.uORF1 (columns 7, 8 and 9) or leaky-scan uAUG-1 and translate *GCN4-lacZ* instead (columns 4, 5 and 6) were calculated from results in columns 1, 2 and 3 by comparing the amount of expression observed for each uORF-containing reporter to the uORF-less construct. Statistically significant differences between mutant and WT are marked with asterisks (\*p\<0.05; \*\*p\<0.005; \*\*\*p\<0.0005).\
10.7554/eLife.31250.013Figure 4---source data 1.Source data for [Figure 4](#fig4){ref-type="fig"} and [Figure 4---figure supplements 1](#fig4s1){ref-type="fig"}, [2](#fig4s2){ref-type="fig"} and [3](#fig4s3){ref-type="fig"}.Effects of substitutions in eIF1A NTT basic residues on eIF1 levels, expression of *SUI1-lacZ* fusions with the native, poor or optimum AUG contexts and eIF1-R13P or R14A substitutions on leaky scanning of el.uORF1 in *GCN4-lacZ* reporters. Effects of selected eIF1A NTT variants overexpression on eIF1 levels, *SUI1-lacZ* fusions expression and *HIS4-lacZ* UUG:AUG expression ratios or *GCN4-lacZ* expression in *SUI3-2* cells. Effects of eIF1-K16A, K16D and overexpression of eIF1A-ΔG8ΔG9 and K10D substitutions on leaky scanning of el.uORF1 in *GCN4-lacZ* reporters.](elife-31250-fig4){#fig4}
Several of the eIF1A variants were expressed at lower than WT levels, including K7A, K7D, K10D, and ΔG8ΔG9 ([Figure 4A](#fig4){ref-type="fig"}), as noted above for UM substitutions K4D, T6D, T6R, and ΔG8 ([Figure 2C](#fig2){ref-type="fig"}). To determine whether the reduced eIF1A expression contributed to their mutant phenotypes, we expressed the subset of variants with the strongest phenotypes from high-copy (hc) plasmids and re-examined their effects on initiation fidelity. The mutant proteins K4D, ΔG8, ΔG8ΔG9 and K10D were expressed from hc plasmids at levels exceeding that of WT eIF1A expressed from a single-copy plasmid (scWT); however, they all still conferred reduced levels of eIF1 expression compared to cells containing normal (scWT) or overexpressed levels of WT eIF1A (hcWT) ([Figure 4---figure supplement 1A](#fig4s1){ref-type="fig"}). The overexpressed variants also conferred reduced expression of the *SUI1-lacZ* fusions with native or poor context ([Figure 4---figure supplement 1B](#fig4s1){ref-type="fig"}); and they co-suppressed the Sui^-^/His^+^ phenotype, elevated UUG:AUG ratio and derepressed *GCN4-lacZ* expression conferred by *SUI3-2* ([Figure 4---figure supplement 2](#fig4s2){ref-type="fig"}). We conclude that the reduced expression of eIF1A NTT variants has little impact on their ability to increase discrimination against poor initiation sites in vivo.
To obtain additional support for the conclusion that eIF1A NTT substitutions increase discrimination against AUGs in poor context, we assayed their effects on *GCN4-lacZ* reporters containing a modified upstream ORF1 elongated to overlap the *GCN4* ORF (el.uORF1). In budding yeast, adenines are preferred at positions −1 to −3 upstream of the AUG (numbered +1 to +3), with an extremely high occurrence of A and low occurrence of C/U at −3; whereas U is preferred at +4 ([@bib61]; [@bib69]). With the WT preferred context of A~-3~A~-2~A~-1~ present at el.uORF1, virtually all scanning ribosomes recognize this AUG codon (uAUG-1), and because reinitiation at the *GCN4* ORF downstream is nearly non-existent, *GCN4-lacZ* expression is extremely low ([@bib18]). Consistent with previous results ([@bib64]), replacing only the optimal A~-3~ with U increases the bypass (leaky scanning) of uAUG-1 to produce an \~15 fold increase in *GCN4-lacZ* translation, whereas introducing the poor context U~-3~U~-3~U~-1~ further increases leaky scanning for a \~50 fold increase in *GCN4-lacZ* expression. Eliminating uAUG-1 altogether increases *GCN4-lacZ* expression by \>100 fold ([Figure 4C](#fig4){ref-type="fig"}, column 1, rows 1--4). From these results, the percentages of scanning ribosomes that either translate el.uORF1 or leaky-scan uAUG-1 and translate *GCN4-lacZ* instead can be calculated ([Figure 4C](#fig4){ref-type="fig"}, cols. 4 and 7; see legend for details), revealing that \>99%, \~89%, and \~66% of scanning ribosomes recognize uAUG-1 in optimum, weak, and poor context, respectively, in WT cells ([Figure 4C](#fig4){ref-type="fig"} col. 7, rows 1--3).
The UM-associated NTT mutation *R13P* increases leaky scanning of uAUG-1, as indicated by increased *GCN4-lacZ* expression for all three reporters containing el.-uORF1 but not for the uORF-less reporter ([Figure 4C](#fig4){ref-type="fig"}, cf cols. 1--2, rows 1--4). Calculating the percentages of ribosomes that recognize uAUG-1 revealed that *R13P* (i) conferred the greatest reduction in recognition of uAUG-1 when the latter resides in poor context, from \~66% to \~27%, (ii) produced a moderate reduction for the weak-context reporter, from \~89% to \~77%, and (iii) evoked only a slight reduction when uAUG-1 is in optimal context, from \>99% to \~98% ([Figure 4C](#fig4){ref-type="fig"}, cf. cols. 7--8, rows 1--3). Similar results were obtained for the targeted mutation *R14A* ([Figure 4C](#fig4){ref-type="fig"}, cf. cols 7 and 9, rows 1--3); and for the targeted *K16A* and *K16D* mutations, with the Asp versus Ala replacement conferring the greater reduction in uAUG-1 recognition ([Figure 4---figure supplement 3A](#fig4s3){ref-type="fig"}, cf. cols. 7--9); and also for the hc*ΔG8ΔG9* and hc*K10D* mutations ([Figure 4---figure supplement 3B](#fig4s3){ref-type="fig"}, cols. 7--9). Thus, both targeted and UM-associated NTT mutations decrease recognition of AUG start codons by scanning PICs preferentially when they reside in poor versus optimum context.
eIF1A NTT substitutions R13P and K16D destabilize the closed, P~IN~ conformation of the 48S PIC in vitro {#s2-4}
--------------------------------------------------------------------------------------------------------
The multiple defects in start codon recognition conferred by the eIF1A NTT mutations suggest that they destabilize the P~IN~ state of the 48S PIC at both UUG and AUG start codons. We tested this hypothesis by analyzing the effects of the R13P and K16D substitutions on the rate of TC dissociation from PICs reconstituted in vitro. Partial 43S·mRNA complexes (lacking eIF3 and eIF5; henceforth p48S PICs) were formed by incubating WT TC (assembled with \[^35^S\]-Met-tRNA~i~ and non-hydrolyzable GTP analog GDPNP) with saturating amounts of eIF1, WT or mutant eIF1A, an uncapped unstructured model mRNA containing either AUG or UUG start codon \[mRNA(AUG) or mRNA(UUG)\], and 40S subunits. p48S PICs containing \[^35^S\]-Met-tRNA~i~ were incubated for increasing time periods in the presence of an excess of unlabeled TC (chase) and then resolved via native gel electrophoresis to separate 40S-bound and unbound fractions of TC. From previous work, it was determined that TC bound in the P~OUT~ state is too unstable to remain associated with the PIC during the native gel electrophoresis used to separate PIC-bound from unbound TC in this assay. It was also deduced that a large proportion of WT complexes formed with mRNA(AUG) achieve a highly stable state from which no TC dissociation occurs during the time-course. A smaller fraction of complexes formed with mRNA (UUG) achieves this highly stable state, and the remainder dissociates with a measurable off-rate ([@bib32]; [@bib13]; [@bib45]).
In agreement with previous findings, little TC dissociation occurred from the WT PICs formed with mRNA(AUG) over the time course of the experiment ([Figure 5A](#fig5){ref-type="fig"}), whereas appreciable dissociation was observed from WT PICS assembled with mRNA(UUG) (k~off~ = 0.18 ± 0.02 h^−1^ ([Figure 5A](#fig5){ref-type="fig"}). Both eIF1A substitutions R13P and K16D increased the extent and rate of TC dissociation from PICs assembled on mRNA(UUG), while having little effect on the mRNA(AUG) complexes ([Figure 5A](#fig5){ref-type="fig"}). The extent of dissociation reflects the proportion of PICs in P~IN~ versus the hyperstable conformation, and the rate of dissociation reflects the stability of the P~IN~ conformation ([@bib32]; [@bib13]). Thus, our results indicate that the eIF1A substitutions R13P and K16D decrease the fraction of the PICs in the hyper-stable conformation and also destabilize the P~IN~ conformation specifically at near-cognate UUG codons. These findings help to account for the decreased utilization of UUG codons (Ssu^-^ phenotype) conferred by these mutations in vivo.
![UM-associated mutant eIF1A-R13P and targeted mutant eIF1A-K16D destabilize the closed/P~IN~ conformation of the 48S PIC at UUG codons in vitro.\
(**A**) Effects of R13P and K16D on TC dissociation kinetics from reconstituted partial 43S•mRNA(AUG) or mRNA(UUG) complexes formed with TC containing \[^35^S\]-Met-tRNA~i~^Met^ and WT eIF1A, eIF1A-R13P or eIF1A-K16D, as indicated. Representative curves are shown for each measurement. Tabulated rate constants (k~off~) and reaction end-points with S.E.M.s are averages of between 4--10 replicate experiments (number in parenthesis); asterisks indicate significant differences between mutant and WT as judged by a two-tailed, unpaired Student's t-test (\*p\<0.05; \*\*p\<0.01). (**B--D**) Effects of R13P and K16D on the dissociation kinetics of fluorescein-labeled eIF1A from reconstituted partial 43S•mRNA complexes, monitored by following changes in fluorescence anisotropy over time after addition of a large excess of unlabeled WT eIF1A. The data for WT eIF1A were fit with a double exponential decay equation, where the fast phase (rate constant k~1~) corresponds to dissociation of eIF1A from the 'open' conformation of the PIC and the second phase (rate constant k~2~) corresponds to dissociation from the 'closed' state ([@bib41]). The ratio of amplitudes of the second phase (α~2~, closed state) to the first phase (α~1,~ open state) is defined as K~amp~. Data for both R13P/K16D were fit to a single exponential equation with rate constant k~1~. R~b~ is the anisotropy of PIC-bound eIF1A. (**B--C**) Representative eIF1A dissociation kinetics from PICs assembled with WT (circles), R13P (squares, panel B), or K16D (triangles, panel C) with mRNA(AUG) (filled symbols) or mRNA(UUG) (open symbols). (**D**) Summary of kinetic parameters from experiments in (**B--C**). Different preparations of labeled WT eIF1A were employed for the experiments examining R13P and K16D, as indicated. All experiments were performed at least two times and errors are average deviations.\
10.7554/eLife.31250.016Figure 5---source data 1.Source data for [Figure 5](#fig5){ref-type="fig"} and [Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}.Effects of eIF1A-R13P and K16D on TC dissociation kinetics from partial 43S•mRNA complexes: rate constants (k~off~) and reaction end-points, and on the dissociation kinetics of fluorescein-labeled eIF1A from partial 43S•mRNA complexes reconstituted with WT eIF2 or the eIF2β-S264Y variant of eIF2: rate constants (k~1~ and k~2~), amplitudes (α~1~ and α~2~,), ratio of amplitudes (K~amp~) and anisotropy of PIC-bound eIF1A (R~b~).](elife-31250-fig5){#fig5}
We also examined the effects of the eIF1A R13P and K16D substitutions on PIC conformation by measuring their effects on the stability of eIF1A binding to the complex. PICs assembled with C-terminally fluorescently-labeled eIF1A were challenged with excess unlabeled eIF1A and the dissociation of labeled eIF1A was monitored over time by following the change in fluorescence anisotropy. The rate of dissociation is generally slow and not physiologically relevant, but reflects the ratio of open to closed PIC conformations and the overall stability of each state ([@bib41]; [@bib15]). The dissociation kinetics are usually biphasic, with the fast and slow phases representing dissociation from the open and closed states, respectively; and the magnitude of each rate constant (k~1~ and k~2~, respectively) reflects the summation of multiple contacts of eIF1A with the PIC. The ratio of amplitudes of the slow phase to the fast phase (K~amp~ = α2/α1) provides an apparent equilibrium constant between the two states, with K~amp~values \> 1 indicating a more prevalent closed state. The anisotropy of the labeled eIF1A in the PIC (R~b~) indicates rotational freedom of the eIF1A CTT, with a higher value indicating greater constraint, which characterizes the closed state.
As expected, WT eIF1A dissociates more slowly from PICs reconstituted with mRNA(AUG) versus mRNA(UUG) ([Figure 5B--C](#fig5){ref-type="fig"}, WT) with both smaller k~2~ and larger K~amp~ values, indicating relatively greater occupancy and increased stability of the closed state at AUG. Consistently, the R~b~ value is greater for mRNA(AUG) versus mRNA(UUG) ([Figure 5D](#fig5){ref-type="fig"}) (Different batches of fluorescein were employed in labeling matched WT and R13P versus WT and K16D proteins, resulting in different R~b~ values for the two WT samples). Both the R13P and K16D substitutions dramatically increase the rate of eIF1A dissociation for both mRNAs ([Figure 5B--C](#fig5){ref-type="fig"}), and evoke monophasic dissociation kinetics with rate constants (k~1~) much greater than the WT k~2~ values for both mRNA(AUG) and mRNA(UUG) ([Figure 5B--D](#fig5){ref-type="fig"}). The R~b~ values also were reduced by both R13P and K16D using either mRNA. These results indicate that both eIF1A NTT substitutions dramatically destabilize the closed conformation of the PIC at both AUG or UUG start codons.
Finally, we examined the effects of R13P on eIF1A dissociation kinetics using eIF2 containing the eIF2ß-S264Y variant encoded by *SUI3-2*. In PICs containing mRNA(UUG) and WT eIF1A, eIF2ß-S264Y decreased k~2~ and increased K~amp~ compared to fully WT PICs, indicating greater occupancy and stability of the closed complex at UUG ([Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}, cf. rows 2--3)---which is consistent with the increased UUG initiation (Sui^-^ phenotype) conferred by *SUI3-2* in vivo. Remarkably, both effects of eIF2ß-S264Y on eIF1A dissociation were reversed on replacing WT eIF1A with the R13P variant, and the R~b~ value was also reduced ([Figure 5---figure supplement 1](#fig5s1){ref-type="fig"}, cf. rows 3--4). These findings help to account for the decreased initiation at UUG codons (Ssu^-^ phenotype) conferred by the eIF1A R13P substitution in *SUI3-2* cells ([Figure 2B](#fig2){ref-type="fig"}). The destabilization of AUG complexes produced by R13P and K16D in the presence of WT eIF2 ([Figure 5B--D](#fig5){ref-type="fig"}) also helps to explain the increased leaky scanning of AUG codons in poor context evoked by these eIF1A substitutions in otherwise WT cells ([Figures 2C--D](#fig2){ref-type="fig"} and [4A--C](#fig4){ref-type="fig"}, and [Figure 4---figure supplement 3A--B](#fig4s3){ref-type="fig"}).
eIF1A-R13P increases discrimination against start codons with non-preferred Kozak context genome-wide {#s2-5}
-----------------------------------------------------------------------------------------------------
To examine effects of the UM-associated R13P substitution in the yeast translatome, we conducted ribosomal footprint profiling of the *R13P* mutant and isogenic WT strain. This technique entails deep-sequencing of mRNA fragments protected from RNase digestion by translating 80S ribosomes arrested in vitro with cycloheximide (Ribo-seq) in parallel with total mRNA sequencing (RNA-seq) ([@bib27]). The translational efficiency (TE) of each mRNA is calculated for each strain as the ratio of sequencing reads for ribosome-protected fragments (RPFs) to total mRNA fragments and normalized to the median ratio among all mRNAs, which is assigned a value of unity. Thus, it should be understood that all TE values are expressed relative to the median TE in that strain. Both ribosome footprinting and RNA-seq results were highly reproducible between the two biological replicates for each strain (Pearson's R \> 0.99) ([Figure 6---figure supplement 1A--D](#fig6s1){ref-type="fig"}). In accordance with the reduced expression of eIF1 conferred by *R13P* ([Figure 2C](#fig2){ref-type="fig"}, eIF1), both RPF and mRNA reads across the *SUI1* coding sequences (CDS) were diminished in *R13P* cells ([Figure 6A](#fig6){ref-type="fig"}). Consistent with these results, we showed previously that the reduced translation of *SUI1* mRNA in eIF1 Ssu^-^ mutants evoked by diminished recognition of its poor-context AUG codon is accompanied by reduced *SUI1* mRNA abundance ([@bib43]). Examples of three other genes with poor context exhibiting reduced translation in *R13P* cells are shown in [Figure 6---figure supplement 2A--C](#fig6s2){ref-type="fig"}, which in one case (*SKI2*) also is accompanied by reduced mRNA levels.
{#fig6}
To determine whether *R13P* evokes widespread discrimination against AUG codons in poor context, we calculated the changes in TE in mutant versus WT cells as the ratio of TE*~R13P~* to TE~WT~ (abbreviated ∆TE*~R13P~*) for groups of genes with different Kozak context. Interestingly, *R13P* conferred a moderate, but significant reduction in TE (log~2~∆TE*~R13P~*\<0) for a group of 123 genes that contain non-preferred bases at every position surrounding the AUG codon, that is (C/U/G)~-3~(C/U/G)~-2~(C/U/G)~-1~(aug)(C/A)~+4~, ([@bib61]) ([@bib69]) that were selected from a set of 4280 genes with adequate read-depth and annotated 5'UTR lengths of ≥5 nt ([Figure 6B](#fig6){ref-type="fig"}, 'Poor' context vs 'All'). The same was true for a larger group of 743 genes containing the least preferred bases C/U at the critical −3 position regardless of the sequence at other positions ([Figure 6B](#fig6){ref-type="fig"}, '−3 C/U' vs 'All'). By contrast, for 200 genes with the preferred bases at all positions, ie. A~-3~A~-2~A~-1~(AUG)(G/U)~+4~, designated 'Perfect' context, or for 3537 genes with A/G at −3, we observed a modest increase in ∆TE*~R13P~* values, compared to all genes ([Figure 6B](#fig6){ref-type="fig"}, 'Perfect', '−3A/G' vs. 'All'). Knowing that changes in *SUI1* mRNA translation are associated with changes in mRNA abundance in the same direction, we repeated the analysis in [Figure 6B](#fig6){ref-type="fig"} considering changes in RPFs rather than TE in the mutant cells, and obtained essentially identical results ([Figure 6---figure supplement 2D](#fig6s2){ref-type="fig"}). These findings indicate that *R13P* increases discrimination against AUG start codons with non-preferred Kozak context at many genes in the manner observed for the *SUI1* AUG ([Figure 2C--D](#fig2){ref-type="fig"}), while conferring an increase in TE for mRNAs with preferred context. Examples of genes exhibiting a relative increase in translation in *R13P* cells are presented in [Figure 6---figure supplement 3](#fig6s3){ref-type="fig"} (panels A-C).
As an orthogonal approach to detecting increased discrimination against poor context by the *R13P* mutation, we sorted genes on the magnitude of ∆TE*~R13P~* values to identify two subsets of genes exhibiting the greatest increases or decreases in TE in mutant cells. As shown in the heat-map of [Figure 6C](#fig6){ref-type="fig"}, there are widespread decreases or increases in TE in *R13P* versus WT cells involving thousands of genes, spanning an \~23 fold range of TE~WT~/TE*~R13P~* ratios from 0.16 to 3.73. We focused on the 100 genes showing the greatest decreases or increases in TE in the mutant versus WT (demarcated with boxes at the top and bottom of the heat-map in [Figure 6C](#fig6){ref-type="fig"}, respectively). The median TE values of these two groups, designated 'TE_down' and 'TE_up', differ significantly between mutant and WT cells ([Figure 6C](#fig6){ref-type="fig"}, boxplots). Constructing sequence logos for positions −6 to −1 and +4 to +6 for these groups of genes revealed that TE_up genes exhibit sequence preferences highly similar to the consensus sequence observed for all 4280 genes ([Figure 6D](#fig6){ref-type="fig"}, TE_up vs. All genes), whereas the TE_down genes lack the strong preference for A/G at −3, as well as the moderate preferences for A at −5 and −2 exhibited by the TE_up group of genes ([Figure 6D](#fig6){ref-type="fig"}, TE_down vs TE_up). We then calculated the AUG context adaptation scores for these sets of genes ([@bib46]), quantifying the similarity between the context of each gene to that of the 2% of all yeast genes with the highest ribosomal load ([@bib69]). Context scores among all yeast genes range from \~0.16 (poorest) to \~0.97 (best), with the most highly expressed genes in yeast exhibiting scores near the top of this range. The 100 genes in the TE_down group exhibit context scores significantly below the median score of \~0.47 for all genes, whereas the context scores for genes in the TE_up group do not differ significantly from the genome-average score ([Figure 6E](#fig6){ref-type="fig"}). Finally, comparing ∆TE*~R13P~* values for 10 deciles of all 4280 genes divided into bins of equal size according to context scores revealed a continuous decline in ∆TE*~R13P~* progressing from bins with the highest to lowest context scores ([Figure 6F](#fig6){ref-type="fig"}).
The correlation between the TE changes conferred by *R13P* and AUG context score shown in [Figure 6E--F](#fig6){ref-type="fig"} was identified without taking into account whether the genes exhibit statistically significant differences in TE between mutant and WT cells. Because too few such mRNAs were identified for rigorous analysis of the correlation, we examined two groups of \~150--200 genes exhibiting significant changes in ribosome occupancy across the CDS between mutant and WT cells (FDR \< 0.1). The 159 genes showing a decrease in ribosome occupancy in *R13P* cells exhibit significantly lower context scores, whereas 214 genes with elevated ribosome occupancies display higher context scores, compared to all 4307 genes examined ([Figure 6---figure supplement 2E](#fig6s2){ref-type="fig"}).
Together, the results indicate that genes with AUG codons in poor context tend to exhibit reductions in TE in *R13P* cells throughout the yeast translatome. The increases in TE observed for genes with preferred context in the mutant might result from decreased competition for limiting initiation factors or 40S subunits owing to reduced translation of mRNAs with poor context. Alternatively, it might partially reflect the normalization of total RPFs and mRNA reads between mutant and WT cells, which sets the average TE value to unity in each strain, such that decreases in TE for one group of genes is necessarily matched by increases in TE for other genes.
We asked next whether changes in TE (or RPFs) conferred by *R13P* might involve other features of the initiation region, including its propensity for forming secondary structures or proximity of the AUG codon to the 5' end of the mRNA---both parameters associated with reduced initiation efficiency in WT cells ([@bib33]; [@bib31]; [@bib22]; [@bib10]). To examine the possible contribution of structure, we interrogated a published database ([@bib31]) wherein each transcribed nucleotide in 3000 different yeast transcripts was assigned a '[p]{.ul}arallel [a]{.ul}nalysis of [R]{.ul}NA [s]{.ul}tructure' (PARS) score, based on its susceptibility in mRNA reannealed in vitro to digestion with nucleases specific for single-stranded or double-stranded RNA, with a higher PARS score denoting a higher probability of double-stranded conformation. For each transcript, we tabulated the average PARS score over the entire 5\'UTR (Average PARS), the sum of PARS scores for the 30nt surrounding the start codon (for genes with a 5\' UTR of ≥16 nt (dubbed 'Start30 PARS'), and the sum of PARS scores for the 30nt centered on the +15 (Plus15) or +30 nucleotides (Plus30) downstream of the AUG. A heat-map depiction of these PARS scores, as well as 5'UTR length, in relation to ∆TE*~R13P~* changes for all 2355 genes with sufficient read-density tabulated in the PARS database revealed no obvious correlation between the magnitude of TE changes conferred by *R13P* and either 5'UTR length or PARS scores ([Figure 6---figure supplement 4A](#fig6s4){ref-type="fig"}). Supporting this, we found no significant difference in the ∆TE*~R13P~* values for a group of 90 mRNAs containing 5' UTR lengths less than ≤18 nt versus all 5136 genes with annotated 5'UTR lengths ([Figure 6---figure supplement 4B](#fig6s4){ref-type="fig"}); and no difference in ∆TE*~R13P~* values between the 1^st^ and 10^th^ deciles of genes binned according to the Start30 or Plus15 PARS scores, representing the two extremes of these PARS scores among all genes ([Figure 6---figure supplement 4C--D](#fig6s4){ref-type="fig"}). These results contrast with our previous findings that genes exhibiting reduced TE on inactivation of RNA helicase Ded1 tend to have unusually long and structured 5'UTRs with greater than average PARS scores ([@bib59]).
We showed above that the *R13P* mutation decreases translation of the elongated version of *GCN4* uORF1 specifically when the uORF1 AUG codon resides in poor context, increasing translation of the downstream CDS of the *GCN4-lacZ* reporter. Hence, we examined our ribosome profiling data for evidence of widespread changes in translation of AUG-initiated uORFs that is dictated by the sequence context of the uORF start codon. It is known that translation of *CPA1* mRNA, encoding an enzyme of arginine biosynthesis, is repressed by its uORF in arginine-replete cells owing to increased pausing during termination at the uORF stop codon, which attenuates progression of scanning PICs to the main *CPA1* AUG codon and elicits nonsense-mediated degradation (NMD) of the transcript ([@bib65]; [@bib16]). The AUG codon of the *CPA1* uORF exhibits a suboptimal context at the −3 and −1 positions, U~-3~A~-2~U~-1~(aug)U~+4~, which is thought to ensure that a fraction of scanning PICs can bypass the uORF and translate *CPA1* at low arginine levels ([@bib65])([@bib16]). Interestingly, *R13P* increases ribosome occupancy in the CDS by \~60%, while decreasing ribosome occupancy in the uORF by \~10%, for a change in uORF relative to CDS ribosome occupancy (designated as relative ribosome occupancy, 'RRO') of 0.58 ([Figure 7A](#fig7){ref-type="fig"}), which suggests diminished recognition of the poor-context uORF AUG and attendant increase in CDS translation. An even greater redistribution of ribosomes from uORF to downstream CDS is illustrated for *ICY1* and *BZZ1,* whose uORF AUG codons depart from optimal context at three or all four positions ([Figure 6---figure supplement 3D--E](#fig6s3){ref-type="fig"}).
{#fig7}
Using bioinformatics, we identified 96 uORFs with AUG start codons that showed evidence of translation in one or more ribosome profiling datasets from WT or various mutant strains, which were obtained in our own laboratory or published by others (see Methods), and which displayed sufficient ribosome occupancies in both the WT and *R13P* strains analyzed here for quantitative analysis. Interestingly, the 72 genes containing uORFs in this group that harbor non-preferred C or U bases at the −3 position mimicked *CPA1* and *ICY1* in showing decreased RRO values in *R13P* versus WT cells, compared to the 24 genes with uORFs containing the preferred bases A or G at −3 ([Figure 7B](#fig7){ref-type="fig"}). Determination of AUG context logos revealed that the base frequencies for the entire group of 96 uORFs differ markedly from that of AUG codons for main CDSs, exhibiting a preference for U/C versus A/G at −3 and little or no preference at the other positions surrounding the uORF ATG (cf. 'All' in [Figure 7C](#fig7){ref-type="fig"} vs 'All' in [Figure 6D](#fig6){ref-type="fig"}), which presumably reflects a need for leaky-scanning of the uORFs. Interestingly, the preference for non-optimal U/C at −3 is even greater, and A is the least prevalent base at −3 for the group of 30 uORFs showing the greatest reductions in RRO in *R13P* cells ([Figure 7C](#fig7){ref-type="fig"}, RRO_down), which is consistent with increased discrimination against uORF AUGs in poor context in the mutant. By contrast, the preference for non-optimal U/C at −3 is eliminated for the 30 uORFs that exhibit the greatest increases in RRO in *R13P* cells (RRO_up), indicating higher frequencies of the preferred A/G bases at this position for this group of uORFs, which is consistent with decreased discrimination in the mutant against uORF AUGs containing relatively stronger sequence contexts ([Figure 7C](#fig7){ref-type="fig"}, RRO_up).
Finally, examination of the AUG context scores for all 96 AUG uORFs reveals a much smaller median score (\~0.22) ([Figure 7D](#fig7){ref-type="fig"}) compared to AUGs for all main CDS (\~0.47; [Figure 6E](#fig6){ref-type="fig"}), supporting our conclusion that AUG uORFs as a group exhibit poor context in order to enable leaky scanning in WT cells. Comparing the context scores between two groups of 30 genes exhibiting the greatest increase in RRO (RRO_up) versus the largest decrease in RRO (RRO_down) in the *R13P* versus WT cells supports the tendency for reduced uORF translation in the mutant when the uORF AUG codon is in poor context but increased uORF translation when the uORF AUG has favorable context ([Figure 7D](#fig7){ref-type="fig"}). Thus, discrimination against suboptimal context contributes to reduced uORF translation, as well as reduced main CDS translation, in *R13P* cells.
The *R13P* mutation increases discrimination against UUG codons in *SUI3-2* and *SUI5* cells ([Figure 2A--B](#fig2){ref-type="fig"}). We found that in cells lacking a Sui^-^ mutation, *R13P* reduced the *HIS4-lacZ* UUG:AUG initiation ratio by a factor of \~2 (from 0.021 ± 0.002 to 0.011 ± 0.001), smaller than the \~3 fold decrease observed in cells containing *SUI3-2* ([Figure 2B](#fig2){ref-type="fig"}). Similarly, we found evidence that *R13P* decreases utilization of the near-cognate ACG start codon that initiates the longer, mitochondrial isoform of alanyl-tRNA synthetase encoded by *ALA1*, reducing the ratio of ribosome occupancies in the N-terminal extension relative to the CDS (ΔNTD/CDS) in the mutant to 0.67 of the WT value ([Figure 7E](#fig7){ref-type="fig"}). This finding is consistent with relaxed discrimination against this native, near-cognate start codon in *R13P* cells.
Discussion {#s3}
==========
In this report we show that all seven substitutions in the NTT of yeast eIF1A associated with uveal melanoma in humans confer hyperaccuracy phenotypes in yeast cells. They suppress inappropriate initiation at a UUG start codon in *his4-301* mRNA to prevent growth in the absence of a histidine supplement. They also reduce the UUG: AUG initiation ratio of a *HIS4-lacZ* reporter, in cells harboring the Sui^-^ mutation in eIF2ß (*SUI3-2*) that reduces accuracy and elevates UUG initiation. Like previously identified Ssu^-^ substitutions in eIF1 ([@bib43]), these eIF1A NTT substitutions also suppress the toxicity of *SUI5* to cell growth at elevated temperatures. Moreover, they decrease initiation at the AUG codons of both *SUI1* mRNA (encoding eIF1) and the *GCN4* uORF1 specifically when they reside in unfavorable Kozak context. The recent structure of a yeast partial 48S PIC predicts that the UM-associated substitutions in the C-terminal portion of the NTT affect direct contacts of the NTT with mRNA nucleotides adjacent to the AUG codon, or in the anticodon of tRNA~i~, and both interactions are thought to stabilize the PIC in the closed conformation with Met-tRNA~i~ accommodated in the P~IN~ state ([@bib26]; [@bib36]). Accordingly, the effects of the UM substitutions in reducing near-cognate UUG and poor-context AUG utilization can be attributed to destabilization of the P~IN~ state with attendant increased requirement for a perfect codon-anticodon duplex and optimal context. Two main lines of evidence support this interpretation. First, an identical set of phenotypes was observed for directed substitutions of conserved basic residues in the NTT that also make direct contacts with mRNA or anticodon nucleotides, namely K7, K10, K16, and R14. Substitutions of these residues with Asp have stronger phenotypes than Ala substitutions, consistent with replacing electrostatic attraction (Lys/Arg) with repulsion (Asp) for the phosphodiester backbone of mRNA or tRNA~i~. The same was true for Asp and Ala substitutions of R13, which is replaced with Pro or His by UM-associated mutations. Second, biochemical experiments reveal that the R13P UM substitution and the directed K16D substitution specifically destabilize the P~IN~ state at UUG codons in vitro, increasing both the fraction of reconstituted PICs from which TC dissociates and the rate of this reaction (k~off~) with a UUG, but not AUG, start codon in the mRNA. These substitutions also increase the rate of eIF1A dissociation, signifying a reduced fraction of PICs in the closed conformation and decreased overall stability of these complexes, with either UUG or AUG start codons. Together, these results help to account for the decreased usage of UUGs and AUGs in poor context conferred by these mutations in vivo, and support the notion that their hyperaccuracy phenotypes result from reduced occupancy and stability of the closed/P~IN~ state that, in turn, confers a heightened requirement for optimal initiation sites.
Although reduced initiation at near-cognate UUG codons in Sui^-^ mutants (Ssu^-^ phenotype) was reported previously for clustered alanine substitutions of eIF1A NTT residues 7--11, 12--16, and 17--21 ([@bib15]), belonging to the scanning inhibitor element designated SI~1~ ([@bib57]), it was unknown which residues in these three segments are most critical for increasing UUG initiation, nor whether any residues in the 7--11 and 12--16 intervals increase initiation at AUGs in poor context. Our findings establish that all five basic residues conserved between yeast and humans that contact mRNA, the anticodon, or 18S rRNA in the decoding center of the py48S PIC (K7, K10, R13, R14, and K16) are critical for efficient utilization of these suboptimal initiation sites, as is the conserved Gly8-Gly9 turn required for making these key contacts ([Figure 1B--C](#fig1){ref-type="fig"}) ([@bib26]). Accordingly, the Ala substitutions of K7 and K10 generated by the *7--11* mutation, and of R13, R14, and K16 by mutation *12--16* are likely responsible for the Ssu^-^ phenotypes of these multiple-residue substitutions ([@bib15]). Although the *17--21* mutation does not substitute any of the key basic residues, it might impair interactions of the C-terminal section of the eIF1A NTT with PIC components and indirectly prevent the basic residues in the N-terminal portion of the NTT from engaging with the decoding center ([Figure 1B](#fig1){ref-type="fig"}). Finally, our results implicate eIF1A residues K3 and K4 (N4 in humans), also substituted in UM, in controlling initiation accuracy, but their molecular functions remain to be determined, as they were not resolved in the py48S structure.
In addition to suppressing the elevated UUG initiation (Sui^-^ phenotype) conferred by the eIF2ß mutation *SUI3-2*, the eIF1A NTT substitutions we analyzed also suppress the derepressed *GCN4-lacZ* expression (Gcd^-^ phenotype) produced by *SUI3-2*. eIF1 stabilizes the open conformation of the PIC, to which TC binds most rapidly (P~OUT~ state) ([Figure 1A](#fig1){ref-type="fig"}) ([@bib52]). The Gcd^-^ phenotypes conferred by other Sui^-^ mutations affecting eIF1, eIF1A, and tRNA~i~ have been attributed to destabilization of this P~OUT~ state of TC binding. This interpretation was based partly on the finding that they are suppressed by Ssu^-^ substitutions in the SI~1~ and SI~2~ elements of eIF1A that destabilize the closed/P~IN~ conformation and thus shift the system in the opposite direction towards the open/P~OUT~ state, which should accelerate TC binding ([@bib57]) ([@bib13]; [@bib45]). Destabilization of the open/P~OUT~ state probably also contributes to the Gcd^-^ phenotype of *SUI3-2* because it is suppressed by Ssu^-^ substitutions in SI~1~ and SI~2~ of eIF1A ([@bib57]). Thus, the marked co-suppression of the Sui^-^ and Gcd^-^ phenotypes of *SUI3-2* observed here for substitutions of the key basic residues K7, K10, R13, and K16 of the NTT, particularly for the acidic Asp replacements, provides additional genetic evidence that they preferentially destabilize the closed/P~IN~ state and shift the system towards the open conformation to which TC loads during assembly of scanning PICs.
We used ribosome footprint profiling to demonstrate that the R13P UM substitution confers a broad decrease in utilization of AUG codons with poor Kozak context in the yeast translatome, mimicking the effect of R13P in reducing eIF1 synthesis from *SUI1* mRNA. R13P also reduced recognition of a subset of the \~100 uORFs whose translation we detected in both mutant and WT cells when their AUG codons reside in poor context, mimicking the effect of R13P of increasing leaky scanning through the elongated version of *GCN4* uORF1 specifically when its AUG codon resides in poor context. *R13P* cells also display somewhat increased discrimination against the near-cognate ACG start codon of the *ALA1* mRNA that initiates an N-terminal extension of the encoded alanyl tRNA synthetase, decreasing the ratio of reads in the extension versus the CDS by \~1/3^rd^. *ALA1* is one of only two annotated genes with non-AUG start codons in yeast ([@bib8]; [@bib7]), and the other such gene, *GRS1,* showed no reduction in initiation from the UUG codon initiating the N-terminal extension of glycyl-tRNA synthetase. This different behavior might be explained by the fact that the context of the *GRS1* UUG matches closely the optimum consensus sequence in yeast in containing A's at −4 to −1, and U at +4, whereas the *ALA1* ACG deviates from this consensus by lacking A's at −1 and −4 and containing A at +4. On the other hand, *R13P* modestly decreased initiation at the UUG codon of the *HIS4-lacZ* reporter, even though it contains preferred A's at −4,--3, and −1. Thus, it remains to be seen whether poor context will be a significant determinant of increased usage of near-cognate start codons conferred by eIF1A NTT Ssu^-^ substitutions.
Considering that the sequence of the yeast and human eIF1A-NTT are quite similar, and that R13 is conserved between the two species ([Figure 1C](#fig1){ref-type="fig"}), our findings for the UM substitutions in yeast eIF1A lead us to propose that the corresponding substitutions in the human eIF1A NTT will favor oncogenic transformation by increasing discrimination against AUG codons with poor context or near-cognate start codons. If one or more tumor suppressor genes contains such a poor initiation site, the UM substitutions can be expected to increase its relative translation rate and thereby impair one or more control mechanisms governing cell proliferation. A recent study on UM tumor cell lines expressing either WT or the G6D variant of EIF1AX provided evidence that the G6D substitution reduces the critical function of EIF1AX in bulk translation initiation. Interestingly, sequencing of total polysomal mRNA indicated that ribosomal protein genes (RPGs) as a group have a heightened requirement for EIF1AX and that the translation of these mRNAs is reduced in G6D vs WT cells ([@bib30]). Given their high rates of translation during rapid cell growth, it seems likely that RPGs would exhibit favorable Kozak context, and by analogy with our findings in yeast on the eIF1A R13P substitution, the RPGs would not be expected to show decreased translation as the result of heightened discrimination against poor context. However, the yeast equivalent of G6D, T6D, did not significantly increase discrimination against the suboptimal eIF1 AUG codon in yeast in the manner observed for R13P. Moreover, unlike G6D in the tumor cells, we found no evidence that the UM-related substitutions in yeast eIF1A reduce bulk initiation. Thus, it is possible that the reduction in RPG expression in G6D tumor cells is a response to reduced bulk translation and cell growth; and it will be interesting to determine whether the R13P substitution in EIF1AX increases discrimination against AUGs in poor context in human cells.
Materials and methods {#s4}
=====================
Plasmid constructions {#s4-1}
---------------------
Plasmids employed in this work are listed in [Table 1](#table1){ref-type="table"}. *TIF11* mutations in plasmids p5633, p5635, p5637, p5638, p5640, p5642 and p5644 were introduced in plasmid p3990 using GeneArtSite-Directed Mutagenesis System (Invitrogen, ThermoFisher) and the appropriate set of complementary mutagenic oligonucleotide primers listed in Table S1, [Supplementary file 1](#supp1){ref-type="supplementary-material"}, following the manufacturer's instructions except for the use of Phusion High fidelity Polymerase (New England BioLabs) for the first step of amplification. Plasmids pDH468, pDH469, pDH481, pDH471, pDH470, pDH473, pDH472, pDH475, pDH474, pDH478, and pDH476 were derived from p3390 by site-directed mutagenesis using the QuickChange XL kit (Agilent Technologies) and the appropriate primers in Table S1. Plasmids pPMB167 to pPMB170 were created by inserting a \~1.2 kb EcoRI-SalI fragment containing *tif11-K4D, tif11-ΔG8, tif11-ΔG8ΔG9* and *tif11-K10D* alleles from p5635, p5640, pDH481 and pDH470, respectively, into the corresponding sites of YCplac181. Plasmids p6013 (*tif11-R13P*) and p6015 (*tif11-K16D*) for expression of eIF1A variants for biochemical analyses were made by PCR amplification of the appropriate DNA fragments from plasmids p5642 and pDH476, respectively and insertion of the resulting fragments into the NdeI-XmaI sites of pTYB2. All plasmids were sequenced to verify the presence of the intended mutations.
10.7554/eLife.31250.023
###### Plasmids used in this study
Plasmid Description Source or reference
----------------------- ------------------------------------------------------------------------------ ---------------------
YCplac111 sc *LEU2* cloning vector ([@bib17])
YEplac181 hc *LEU2* cloning vector ([@bib17])
YCplac22 sc *TRP1* cloning vector ([@bib17])
p3390/pDSO9 sc *LEU2 TIF11* in YCplac111 ([@bib9])
p5633 sc *LEU2 tif11-K3E* in YCplac111 This study
p5635 sc *LEU2 tif11-K4D* in YCplac111 This study
p5638 sc *LEU2 tif11-T6D* in YCplac111 This study
p5637 sc *LEU2 tif11-T6R* in YCplac111 This study
p5640 sc *LEU2 tif11-ΔG8* in YCplac111 This study
p5642 sc *LEU2 tif11-R13P* in YCplac111 This study
p5644 sc *LEU2 tif11-G15D* in YCplac111 This study
pDH469 sc *LEU2 tif11-K7A* in YCplac111 This study
pDH468 sc *LEU2 tif11- K7D* in YCplac111 This study
pDH481 sc *LEU2 tif11-ΔG8ΔG9* in YCplac111 This study
pDH471 sc *LEU2 tif11-K10A* in YCplac111 This study
pDH470 sc *LEU2 tif11-K10D* in YCplac111 This study
pDH473 sc *LEU2 tif11-R13A* in YCplac111 This study
pDH472 sc *LEU2 tif11-R13D* in YCplac111 This study
pDH475 sc *LEU2 tif11-R14A* in YCplac111 This study
pDH474 sc *LEU2 tif11-R14D* in YCplac111 This study
pDH478 sc *LEU2 tif11-K16A* in YCplac111 This study
pDH476 sc *LEU2 tif11-K16D* in YCplac111 This study
p3400/pDSO23 hc *LEU2 TIF11* in YEplac181 ([@bib9])
pPMB167 hc *LEU2 tif11-K4D* in YEplac181 This study
pPMB168 hc *LEU2 tif11-ΔG8* in YEplac181 This study
pPMB169 hc *LEU2 tif11-ΔG8ΔG9* in YEplac181 This study
pPMB170 hc *LEU2 tif11- K10D* in YEplac181 This study
p4281/YCpTIF5-G31R-W sc *TRP1 TIF5-G31R* in YCplac22 ([@bib63])
p4280/YCpSUI3-S264Y-W sc *TRP1 SUI3-S264Y* in YCplac22 ([@bib63])
p367 sc *URA3 HIS4(ATG)-lacZ* ([@bib11])
p391 sc *URA3 HIS4(TTG)-lacZ* ([@bib11])
p180 sc *URA3 GCN4-lacZ* ([@bib20])
pPMB24 sc *URA3 SUI1-lacZ* ([@bib43])
pPMB25 sc *URA3 SUI1-opt-lacZ* ([@bib43])
pPMB28 sc *URA3 SUI1~UUU~-lacZ* ([@bib43])
pC3502 sc *URA3* ^-3^AAA^−1^ el.uORF1 *GCN4-lacZ* in YCp50 ([@bib64])
pC4466 sc *URA3* ^-3^UAA^−1^ el.uORF1 *GCN4-lacZ* in YCp50 ([@bib64])
pC3503 sc *URA3* ^-3^UUU^−1^ el.uORF1 *GCN4-lacZ* in YCp50 ([@bib64])
pC3505 sc *URA3* el.uORF1-less *GCN4-lacZ* in YCp50 ([@bib64])
pTYB2-TIF11 *TIF11* in pTYB2 ([@bib1])
p6013 *tif11-R13P* in pTYB2 This study
p6015 *tif11-K16D in pTYB2* This study
pRaugFFuug Dual luciferase reporter *LUC~ren~(aug)-LUC~firefly~ (uug)* in *URA3* vector ([@bib32])
pRaugFFuug Dual luciferase reporter *LUC~ren~(aug)-LUC~firefly~ (uug)* in *URA3* vector ([@bib32])
Yeast strain constructions {#s4-2}
--------------------------
Yeast strains employed in this work are listed in [Table 2](#table2){ref-type="table"}. Derivatives of strain H3582 \[*MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3392 (sc *URA3 TIF11*)\] were constructed by transforming H3582 to Leu^+^ with single copy (sc) or high copy (hc) *LEU2* plasmids harboring the appropriate *TIF11* alleles on synthetic complete medium (SC) lacking leucine (SC-Leu), and the resident *TIF11^+^URA3* plasmid (p3392) was evicted by selecting for growth on 5-FOA medium. Derivatives of strain H3582 containing plasmid-borne *SUI5* (p4281/YCpTIF5-G31R-W), *SUI3-2* (p4280/YCpSUI3-S264Y-W) or empty vector were generated by transformation and selection on SC lacking leucine and tryptophan (SC-Leu-Trp). Strains FZY010/FZY011 and PMY337/PMY338 used for ribosome profiling are independent transformants of strains PMY290 and PMY318 with *TRP1* vector YCplac22, respectively.
10.7554/eLife.31250.024
###### Yeast strains used in this study
Strain Genotype Source
--------- --------------------------------------------------------------------------------------------------------------------------------------------- ------------
H3582 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3392 (sc *URA3 TIF11*) ([@bib14])
PMY318 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3390 (sc *LEU2 TIF11*) This study
PMY284 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5633 (sc *LEU2 tif11-K3E*) This study
PMY285 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5635 (sc *LEU2 tif11-K4D*) This study
PMY286 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5638 (sc *LEU2 tif11-T6D*) This study
PMY287 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5637 (sc *LEU2 tif11-T6R*) This study
PMY289 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5640 (sc *LEU2 tif11-ΔG8*) This study
PMY290 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5642 (sc *LEU2 tif11-R13P*) This study
PMY291 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5644 (sc *LEU2 tif11-G15D*) This study
PMY320 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH469 (sc *LEU2 tif11-K7A*) This study
PMY321 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH468 (sc *LEU2 tif11-K7D*) This study
PMY322 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH481 (sc *LEU2 tif11-ΔG8ΔG9*) This study
PMY323 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH471 (sc *LEU2 tif11-K10A*) This study
PMY324 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH470 (sc *LEU2 tif11-K10D*) This study
PMY325 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH473 (sc *LEU2 tif11-R13A*) This study
PMY326 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH472 (sc *LEU2 tif11-R13D*) This study
PMY327 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH475 (sc *LEU2 tif11-R14A*) This study
PMY329 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH474 (sc *LEU2 tif11-R14D*) This study
PMY330 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH478 (sc *LEU2 tif11-K16A*) This study
PMY332 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH476 (sc *LEU2 tif11-K16D*) This study
PMY354 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3400 (hc *LEU2 TIF11*) This study
PMY355 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB167 (hc *LEU2 tif11-K4D*) This study
PMY357 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB168 (hc *LEU2 tif11-ΔG8*) This study
PMY358 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB169 (hc *LEU2 tif11-ΔG8ΔG9*) This study
PMY359 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB170 (hc *LEU2 tif11- K10D*) This study
PMY32 *MATa ura3-52 leu2-3 leu2-112 trp1Δ−63 his4-301*(*ACG*) *sui1Δ::hisG* pPMB02 (sc *LEU2 sui1-K60E*) ([@bib43])
PMY293 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3390 (sc *LEU2 TIF11*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY295 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5633 (sc *LEU2 tif11-K3E*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY296 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5635 (sc *LEU2 tif11-K4D*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY297 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5638 (sc *LEU2 tif11-T6D*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY298 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5637 (sc *LEU2 tif11-T6R*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY300 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5640 (sc *LEU2 tif11-ΔG8*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY301 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5642 (sc *LEU2 tif11-R13P*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY302 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5644 (sc *LEU2 tif11-G15D*) p4281 (sc *TRP1 TIF5-G31R*) This study
PMY335 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3390 (sc *LEU2 TIF11*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY310 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5633 (sc *LEU2 tif11-K3E*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY311 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5635 (sc *LEU2 tif11-K4D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY312 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5638 (sc *LEU2 tif11-T6D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY313 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5637 (sc *LEU2 tif11-T6R*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY315 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5640 (sc *LEU2 tif11-ΔG8*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY316 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5642 (sc *LEU2 tif11-R13P*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY317 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5644 (sc *LEU2 tif11-G15D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY339 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH469 (sc *LEU2 tif11-K7A*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY340 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH468 (sc *LEU2 tif11-K7D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY341 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH481 (sc *LEU2 tif11-ΔG8ΔG9*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY342 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH471 (sc *LEU2 tif11-K10A*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY343 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH470 (sc *LEU2 tif11-K10D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY344 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH473 (sc *LEU2 tif11-R13A*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY345 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH472 (sc *LEU2 tif11-R13D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY346 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH475 (sc *LEU2 tif11-R14A*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY348 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH474 (sc *LEU2 tif11-R14D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY349 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH478 (sc *LEU2 tif11-K16A*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY351 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pDH476 (sc *LEU2 tif11-K16D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY337 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3390 (sc *LEU2 TIF11*) YCplac22 (sc *TRP1*) This study
PMY338 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3390 (sc *LEU2 TIF11*) YCplac22 (sc *TRP1*) This study
PMY360 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3400 (hc *LEU2 TIF11*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY362 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB167 (hc *LEU2 tif11-K4D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY364 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB168 (hc *LEU2 tif11-ΔG8*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY365 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB169 (hc *LEU2 tif11-ΔG8ΔG9*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY366 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* pPMB170 (hc *LEU2 tif11- K10D*) p4280 (sc *TRP1 SUI3-S264Y*) This study
PMY361 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p3400 (hc *LEU2 TIF11*) YCplac22 (sc *TRP1*) This study
GP3511 *MATα ura3-52 leu2-3 leu2-112 ino1 sui2∆ gcn2∆ pep4::LEU2 \< HIS4 lacZ,ura3−52 \> *pAV1089 (*SUI2,SUI3,GCD11-HIS,URA3*) ([@bib53])
H4560 *MATα ura3-52 leu2-3 leu2-112 ino1 sui2Δ gcn2Δ pep4::leu2::natMX sui3Δ::kanMX \< HIS4 lacZ,ura3−52 \> *p5321 (*SUI2,SUI3-2,GCD11-HIS,LEU2*) ([@bib45])
YAS2488 *MATa leu2-3,−112 his4-53a trp1 ura3-52 cup1::LEU2/PGK1 pG/MFA2* pG ([@bib2])
FZY010 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5642 (sc *LEU2 tif11-R13P*) YCplac22 (sc *TRP1*) This study
FZY011 *MATa ura3-52 trp1Δ63 leu2-3, leu2-112 his4-301(ACG) tif11Δ* p5642 (sc *LEU2 tif11-R13P*) YCplac22 (sc *TRP1*) This study
Biochemical assays using yeast cell extracts {#s4-3}
--------------------------------------------
Assays of β-galactosidase activity in whole cell extracts (WCEs) were performed as described previously ([@bib47]). At least four biological replicates (independent transformants) were employed for all β-galactosidase activity measurements. Unpaired t-tests were performed to compare wild type and mutant mean values and the change was considered significant if the two-tailed P value was \< 0.05. Luminescence expressed from dual luciferase reporter plasmids pRaugFFuug and pRaugFFaug was measured as described previously ([@bib32]). For Western analysis, WCEs from three biological replicates (independent transformants) were prepared by trichloroacetic acid extraction as previously described ([@bib56]) and immunoblot analysis was conducted as previously described ([@bib49]) using antibodies against eIF1A/Tif11 ([@bib51]), eIF1/Sui1 ([@bib63]) and eIF2Bε/Gcd6 ([@bib6]). Two technical replicates were performed using the same extracts and two different amounts of each extract differing by 2-fold were loaded in successive lanes. Enhanced chemiluminiscence (Amersham) was used to visualize immune complexes, and signal intensities were quantified by densitometry using NIH ImageJ software.
Biochemical analysis in the reconstituted yeast system {#s4-4}
------------------------------------------------------
WT eIF1 and eIF1A and eIF1A variants R13P and K16D were expressed in BL21(DE3) Codon Plus cells (Agilent Technologies) and purified using the IMPACT system (New England Biolabs) as described previously ([@bib1]). His~6~-tagged WT eIF2, or the variant containing eIF2β-S264Y, were overexpressed in yeast strains GP3511 and H4560, respectively, and purified as described ([@bib1]). 40S subunits were purified as described previously from strain YAS2488 ([@bib1]). Model mRNAs with sequences 5′-GGAA\[UC\]~7~UAUG\[CU\]~10~C-3′ and 5′-GGAA\[UC\]~7~UUUG\[CU\]~10~C-3′ were purchased from Thermo Scientific. Yeast tRNA~i~^Met^ was synthesized from a hammerhead fusion template using T7 RNA polymerase, charged with \[^35^S\]-methionine, and used to prepare radiolabeled eIF2·GDPNP·\[^35^S\]-Met-tRNA~i~ ternary complexes (\[^35^S\]-TC), all as previously described ([@bib1]). Yeast Met-tRNA~i~^Met^ was purchased from tRNA Probes, LLC and used to prepare unlabeled TC in the same way. For eIF1A dissociation kinetics, the WT and mutant eIF1A proteins were labeled at their C-termini with Cys-Lys-ε-fluorescein dipeptide, using the Expressed Protein Ligation system as previously described ([@bib40]).
### TC and eIF1A dissociation kinetics {#s4-4-1}
TC dissociation rate constants (k~off~) were measured by monitoring the amount of \[^35^S\]-TC that remains bound to 40S·eIF1·eIF1A·mRNA (43S·mRNA) complexes over time, in the presence of excess unlabeled TC (chase), using a native gel shift assay to separate 40S-bound from unbound \[^35^S\]-TC. 43S·mRNA complexes were preassembled for 2 hr at 26°C in reactions containing 40S subunits (20 nM), eIF1 (1 µM), eIF1A (WT or mutant variants, 1 µM), mRNA (10 µM), and \[^35^S\]-TC (0.25 µM eIF2/0.1 mM GDPNP/1 nM \[^35^S\]-Met-tRNA~i~) in 60 µl of reaction buffer (30 mM Hepes-KOH (pH 7.4), 100 mM potassium acetate (pH 7.4), 3 mM magnesium acetate, and 2 mM dithiothreitol). To initiate each dissociation reaction, a 6 µl-aliquot of the preassembled 43S·mRNA complexes was mixed with 3 µl of 3-fold concentrated unlabeled TC chase (comprised of 2 µM eIF2/0.3 mM GDPNP/0.9 µM Met-tRNA~i~), to achieve in the final dissociation reaction a 300-fold excess of unlabeled TC (\~300 nM) over labeled TC (\~1 nM), based on the two different amounts of Met-tRNA~i~ employed, and incubated for the prescribed period of time. A converging time course was employed so that all dissociation reactions are terminated simultaneously by the addition of native-gel dye and loaded directly on a running native gel. The fraction of \[^35^S\]-Met-tRNA~i~ remaining in 43S complexes at each time point was determined by quantifying the 40S-bound and unbound signals using a PhosphorImaging, normalized to the ratio observed at the earliest time-point, and the data were fit with a single exponential equation ([@bib32]).
The kinetics of eIF1A dissociation were determined exactly as described earlier ([@bib58]).
Ribosome footprint profiling and RNA-Seq {#s4-5}
----------------------------------------
Ribosome profiling was conducted essentially as described previously ([@bib60]) as detailed below, on isogenic strains FZY010 and FZY011 (*tif11-R13P*), and PMY337 and PMY338 (WT *TIF11*), providing two biological replicates of each genotype, cultured in SC-Leu-Trp, except that cells were not treated with cycloheximide before harvesting, and cycloheximide was added to the lysis buffer at 5x the standard concentration. In addition, RNAse-trimmed ribosomes were purified by velocity sedimentation through sucrose gradients prior to extraction of mRNA; and Illumina Ribo-Zero Gold rRNA Removal Kit (Yeast) was employed on linker-ligated mRNA in lieu of poly(A) selection. Genes with less than 128 total mRNA reads, or less than 40 total RPF reads, in the four samples combined (two replicates of both WT and mutant strains) were excluded from the calculation of TE values.
### Generation, processing, and analysis of sequence libraries of ribosome protected footprints or total mRNA fragments {#s4-5-1}
*tif11-R13P* (FZY010, FZY011) and WT (PMY337, PMY338) yeast strains growing exponentially in SC medium at 30°C were harvested by vacuum filtration at room temperature, without prior treatment with cycloheximide, and quick-frozen in liquid nitrogen. Cells were lysed in a freezer mill with lysis buffer (20 mM Tris \[pH 8.0\], 140 mM KCl, 1.5 mM MgCl~2~, 1% Triton, 500 µg/mL cycloheximide). For ribosome footprint library preparation, 30 A~260~ units of extract were treated with 450U of RNAse I (Ambion, \#AM2295) for 1 hr at 25°C on a Thermomixer at 700 rpm, and 80S ribosomes were purified by sedimentation through a sucrose density gradient as described ([@bib19]). Ribosome-protected mRNA fragments (footprints) were purified using a miRNeasy Mini kit (Qiagen) per the vendor\'s instructions. Following size selection and dephosphorylation, a Universal miRNA cloning linker (New England Biolabs, \#S1315S) was ligated to the 3' ends of footprints, followed by reverse transcription, circular ligation, rRNA subtraction, PCR amplification of the cDNA library, and DNA sequencing with an Illumina HiSeq system. For RNA-seq library preparation, total RNA was purified using miRNeasy Mini kit (Qiagen) from aliquots of the same extracts used for footprint library preparation, 5 µg total RNA was randomly fragmented at 70°C for 8 min in fragmentation reagent (Ambion \#AM8740). Fragment size selection, library generation and sequencing were carried out as above, except Ribo-Zero Gold rRNA Removal Kit (Yeast) was employed to remove rRNAs after linker-ligation. Linker sequences were trimmed from Illumina reads and the trimmed fasta sequences were aligned to the *S. cerevisiae* ribosomal database using Bowtie ([@bib34]). The non-rRNA reads (unaligned reads) were then mapped to the *S. cerevisiae* genome using TopHat ([@bib62]). Wiggle track normalization for viewing RPF or RNA reads in the IGV browser was conducted as follows. Wiggle files were produced from the alignment file, one each for genes on the Watson or Crick strand. The total reads on both strands were summed and a normalization factor *q* was calculated as 1000,000,000/(total reads on W + C strands). Wiggle files were then regenerated by multiplying all reads by the factor *q*, yielding the number of reads per 1000 million total reads (rpkm). uORFs with evidence of translation in WT and R13P cells were identified as follows. First, we employed the yassour-uorf program of ([@bib5]) that identifies all potential uORFs within annotated 5'UTRs initiating with either AUG or a near-cognate codon and then quantifies the footprints in the +1 and −1 codons of all putative uORFs. A uORF was judged to be translated if the +1 to −1 footprint ratio exceeded four and the total footprint counts at +1 and −1 exceeded 15, and also if the reads in the zero frame are at least 50% of the reads in all three frames (ie. -c15-r4-z0.5 in the relevant line of code). This analysis was conducted on multiple published and unpublished datasets summarized in Table S2, [Supplementary File 1](#supp1){ref-type="supplementary-material"}. After excluding uORFs shorter than three codons, we identified 564 AUG-initiated uORFs and 5497 near-cognate uORFs with evidence of translation in one or more experiments. In the second step, we validated \~51% and \~44% of the AUG uORFs and near-cognate uORFs, respectively, by employing a distinct uORF identification tool, RibORF ([@bib29]), which is based on the criteria of 3-nt periodicity and uniformity of read distribution across uORF codons. Applying a moderately stringent probability of prediction of \>0.5, RibORF confimed that 291 AUG uORFs and 2429 near-cognate uORFs show evidence of translation in the datasets from which they were first identified by the yassour-uorf program. A bed file was generated containing the sequence coordinates of every uORF and combined with a bed file containing the coordinates of the 5'UTR, main CDS, and 3'UTR of each gene, and used to obtain footprint (FP) counts for 5'UTRs, uORFs, and main CDS in each strain examined, excluding the first and last nucleotide triplets of 5'UTRs, the first and last codons of uORFs, and the first 20 codons of CDS. mRNA read counts were determined for all codons of the main CDS. DESEQ ([@bib4]) was employed for differential expression analysis of changes in TE, RPFs, or RRO values, and to impose cutoffs for minimum read numbers (as indicated in figure legends) and remove outliers.
For all notched box-plots, constructed using a web-based tool at <http://shiny.chemgrid.org/boxplotr/>, the upper and lower boxes contain the second and third quartiles and the band gives the median. If the notches in two plots do not overlap, there is roughly 95% confidence that their medians are different.
The AUG context adaptation index (context score) ([@bib46]) was calculated as A~UG~CAI = (w~-6~ x w~-5~ x w~-4~ x w~-3~ x w~-2~ x w~-1~ x w~+1~ x w~+2~ x w~+3~)^1/9^ where w~i~ is the fractional occurrence of that particular base, normalized to the most prevalent base, present in the i^th^ position of the context among the \~270 most highly expressed yeast genes, taken from the matrix of frequencies and relative adaptiveness (w) of the nucleotide in the AUG context of this group of \~270 reference genes ([@bib69]). The context scores range from \~0.16 (poorest) to \~0.97 (best) among all yeast genes.
Accession numbers {#s5}
=================
Sequencing data from this study have been submitted to the NCBI Gene Expression Omnibus (GEO; <http://www.ncbi.nlm.nih.gov/geo/>) and the accession numbers are listed in the Additional Files under Major datasets.
Funding Information
===================
This paper was supported by the following grant:
- http://dx.doi.org/10.13039/100000002National Institutes of Health Intramural Research Program to Alan G Hinnebusch.
We thank Nicholas Ingolia, Nicholas Guydosh, and David Young for protocols and helpful discussions about ribosome profiling data analysis, Swati Gaikwad for help in analysis of AUG context scores and Shardul Kulkarni for sharing ribosome profiling data prior to publication. This work was supported in part by the Intramural Research Program of the National Institutes of Health. PM-M was financed in part by the Spanish Ministry of Economy, Industry and Competitiveness (MINECO) and European FEDER funds, through Project RTC2015-4391-2 awarded to MT.
Additional information {#s6}
======================
Reviewing editor, *eLife*.
No competing interests declared.
Conceptualization, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing---original draft, Writing---review and editing.
Data curation, Formal analysis, Validation, Investigation, Visualization, Writing---original draft, Writing---review and editing.
Investigation.
Formal analysis, Investigation, Writing---original draft, Writing---review and editing.
Investigation, Writing---original draft.
Formal analysis, Investigation, Writing---original draft, Writing---review and editing.
Investigation, Formal analysis.
Resources, Writing---original draft, Writing---review and editing.
Investigation, Writing---original draft, Writing---review and editing.
Supervision, Writing---original draft, Writing---review and editing.
Conceptualization, Resources, Funding acquisition, Validation, Visualization, Methodology, Writing---original draft, Project administration, Writing---review and editing.
Additional files {#s7}
================
10.7554/eLife.31250.025
###### Supplementary Tables.
Table S1: Oligonucleotide primers employed for *TIF11* mutagenesis in this study Table S2: Ribosome profiling datasets used for uORF identification
10.7554/eLife.31250.026
###### Excel file containing results and analyses from ribosome footprint profiling of WT and *tif11-R13P* cells.
Spreadsheet 1, '*CDS\_*all Expr', tabulates log~2~ values of the following parameters for the 5037 expressed genes listed in columns A-B for WT and *tif11-R13P* cells: Ribosome footprint sequencing reads (RPF_WT and RPF\_*R13P*); mRNA sequencing reads (mRNA_WT and mRNA\_*R13P*); the ratios RPF\_*R13P*/RPF_WT (ΔRPF\_ *R13P*) and mRNA\_*R13P/*mRNA_WT (ΔmRNA\_*R13P*); and the ratio ΔRPF*\_R13P/*ΔmRNA*\_R13P* (ΔTE\_*R13P*). Spreadsheet 2, 'Context_score', contains a subset of genes in Spreadsheet 1 (4280 genes) with 5'UTR length \>5 nt examined for additional parameters: 5'UTR length, sequences between positions −6 and +6, and the context scores. Spreadsheet 3, 'AUG_uORFs_identified', contains all 564 AUG uORFs identified using the yassour-uorf program from multiple datasets listed in Table S2, listing uORF chromosome coordinates, distances of the uORF AUG from the 5' end of the mRNA and the main CDS start codon, uORF sequence contexts between the −3 and +4 positions and the context scores (NA, for uORF 5' UTR \<3 nt). Spreadsheet 4, '*uORF_Expr*', tabulates log~2~ values of the following parameters for the 97 expressed uAUG uORFs listed in column A for WT and *tif11-R13P* cells: Ribosome footprint sequencing reads on CDS (RPF_CDS_WT and RPF_CDS\_*R13P*); Ribosome footprint sequencing reads on uORFs (RPF_uORF_WT and RPF_uORF\_*R13P*); the ratios RPF_CDS\_*R13P*/RPF_CDS_WT (ΔRPF_CDS\_ *R13P*) and RPF_uORF\_*R13P*/RPF_uORF_WT (ΔRPF_uORF\_*R13P*); Relative Ribosome Occupancy (RRO) for WT (the ratio of RPF_uORF_WT/RPF_CDS_WT, RRO_WT) and *R13P* (the ratio of RPF_uORF\_*R13P*/RPF_CDS\_*R13P, RRO_R13P*) and the ratio RRO*\_R13P/*RRO*\_WT* (RRO\_*R13P*); Spreadsheet 5, PARS score analysis of 5' UTRs of 2642 genes, conducted as described in ([@bib60]) and their context scores as listed in Spreadsheet 2. Spreadsheet 6 '*CDS\_*RPF_Change', tabulates log~2~ values of the ratio RPF\_*R13P*/RPF_WT (ΔRPF\_*R13P* (log2)) and adjusted p-value (padj) for the 5083 expressed genes detected by the DESEQ2 package listed in columns A-B
10.7554/eLife.31250.027
Major datasets {#s8}
--------------
The following datasets were generated:
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017eIF1A residues implicated in cancer stabilize translation preinitiation complexes and favor suboptimal initiation sites in yeast<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE108334>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSE108334)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for SUI1-L96P)\_with_CHX_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895470>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895470)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for SUI1-L96P)\_with_CHX_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895471>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895471)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI1-L96P_w/o_CHX_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895472>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895472)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI1-L96P_w/o_CHX_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895473>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895473)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI1-L96P_with_CHX_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895468>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895468)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI1-L96P_with_CHX_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895469>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895469)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for SUI1-T15A)\_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895458>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895458)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for SUI1-T15A)\_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895459>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895459
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI1-T15A_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895456>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895456)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI-T15A_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895457>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895457)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2(for SUI1_T15A)\_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895460>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895460)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2(for SUI1_T15A)\_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895461>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895461)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2/SUI-T15A_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895462>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895462)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2/SUI-T15A_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895463>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895463)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_20_WT_with_CHX_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895476>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895476)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for TIF11_R13P)\_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895450>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895450)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for TIF11_R13P)\_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895451>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895451)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_TIF11_R13P_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895448>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895448)
Pilar Martin-MarcosCharm KarunasiriFujun ZhouFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2(for TIF11-R13P)\_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895452>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895452)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2(for TIF11-R13P)\_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895453>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895453)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2/TIF11_R13P_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895454>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895454)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_SUI3-2/TIF11_R13P_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895455>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895455)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for SUI1-L96P)\_w/o_CHX_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895474>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895474)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G HinnebuschFan Zhang2017ribo_WT(for SUI1-L96P)\_w/o_CHX_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895475>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895475
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_37_WT_w/o_CHX_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895491>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895491)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for tif3)\_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895466>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895466)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_WT(for tif3)\_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895467>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895467)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_tif3△\_1<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895464>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895464)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_tif3△\_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895465>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895465)
Pilar Martin-MarcosFujun ZhouCharm KarunasiriFan ZhangJinsheng DongJagpreet NandaShardul D KulkarniNeelam Dabas SenMercedes TamameMichael ZeschnigkJon R LorschAlan G Hinnebusch2017ribo_TIF11_R13P_2<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM2895449>Publicly available at the NCBI Gene Expression Omnibus (accession no. GSM2895449)
The following previously published datasets were used:
Gerashchenko MGladyshev V2014Translation Inhibitors Cause Abnormalities in Ribosome Profiling Experiments<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE59573>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE59573)
Guydosh NRGreen R2013Ribosome profiling study of dom34 and hbs1 knockout strains using short (16-nt) and long (28-nt) monosome-protected footprints and disome-protected footprints<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE52968>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE52968)
Sen NDZhou FIngolia NTHinnebusch AG2015Genome-wide analysis of translational efficiency reveals distinct but overlapping functions of yeast DEAD-box RNA helicases Ded1 and eIF4A<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE66411>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE66411)
Sen NDZhou FHarris MSIngolia NTHinnebusch AG2016eIF4B preferentially stimulates translation of long mRNAs with structured 5\'UTRs and low closed-loop potential but weak dependence on eIF4G<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE81966>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE81966)
Young DJGuydosh NRZhang FHinnebusch AGGreen R2015Ribosome profiling study of rli1 depeletion strain<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE69414>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE69414)
Kertesz MWan YMazor ERinn JLNutter RCChang HYSegal E2010Genome-wide Measurement of RNA Secondary Structure in Yeast<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE22393>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE22393)
Pelechano VWei WSteinmetz LM2013Saccharomyces cerevisiae Transcript Isoform mapping<https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE39128>Publicly available at the NCBI Gene Expression Omnibus (accession no: GSE39128)
10.7554/eLife.31250.105
Decision letter
Sonenberg
Nahum
Reviewing Editor
McGill University
Canada
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
Thank you for submitting your article \"eIF1A residues implicated in cancer stabilize translation preinitiation complexes and favor suboptimal initiation sites\" for consideration by *eLife*. Your article has been favorably evaluated by James Manley (Senior Editor) and three reviewers, one of whom, Nahum Sonenberg (Reviewer \#1), is a member of our Board of Reviewing Editors. The following individual involved in review of your submission has agreed to reveal their identity: Matt Sachs (Reviewer \#3).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Summary:
The authors present a study that provides an extensive analysis of initiation codon selection in response to N-terminal mutants of yeast eIF1A. The study primarily focuses on cancer-associated eIF1A-NTT mutants. Cancer-associated mutants are introduced into the yeast eIF1A protein together with some synthetic yeast mutants in the same region. Mutants confer a hyper-accuracy phenotype that is likely attributable to destabilization of a closed/Pin state of the scanning 48S PIC. The most potent mutant R13P confers altered start site selection genome-wide in yeast, which is consistent with the genetic experiments and in vitro kinetics.
Previous work from Hinnebusch and coworkers has shown that specific residues in eIF1, eIF1A and eIF2β function in the discrimination of poor context AUG recognition and non-AUG codons (Martin-Marcos et al. MCB 2011 Dec; 31(23): 4814-4831; Saini et al. 2010 Genes Dev. 24: 97-110). The current study now provides far more detail into how eIF1A plays a role in start site selection. Interestingly, the finding that essentially all cancer-associated mutations of eIF1A, when made in yeast eIF1A, confer a consistent hyper-accuracy phenotype, is very striking. Overall, the experiments are well undertaken and the data are appropriately interpreted.
Importantly, the authors\' findings indicate that the increased stringency of selection of AUG codons in the optimal Kozak context favors oncogenic transformation. The authors offer the reasonable prediction \"If one or more tumor suppressor genes contains such a poor initiation site, the μm substitutions can be expected to increase its relative translation rate and thereby impair one or more control mechanisms governing cell proliferation\"
Essential revisions:
1\) One concern of the study is whether the cancer-associated mutants in eIF1A are applicable to human eIF1A function and cancer. The eIF1A-NTT is highly conserved between two species, but the Kozak consensus and dependence on it are generally thought to vary between yeast and humans. The study, therefore, provides new insight into how the eIF1A-NTT maintains the fidelity of start codon selection, but the possible role of the cancer-associated residues in human eIF1A in mammalian start site selection is not determined. Ideally, this should be addressed directly, but at a minimum the end of the title should be altered to include \"in yeast\" to better describe the study.
2\) They must have become aware of a very relevant paper Johnson et al. PLoS One. 2017 Jun 8;12(6):e0178189). The authors did RNAseq of polysome fractions prepared from actual μm cancer cells harboring the eIF1A-NTT mutation. They identified genes whose translational efficiency was affected by the knockdown of EIF1AX. Could the authors look for relevant information regarding mRNAs from the list? The paper should be referenced and its implications discussed.
3\) [Figure 5](#fig5){ref-type="fig"} convincingly demonstrates that the R13P and K16D mutations in eIF1A destabilize the closed/Pin conformation of the 48S PIC at UUG codons in vitro. Could they extend these results by showing a decreased UUG/AUG start codon usage in a cell-free translation system from eIF1A-mutant cells as compared to wildtype?
10.7554/eLife.31250.106
Author response
> Essential revisions:
>
> 1\) One concern of the study is whether the cancer-associated mutants in eIF1A are applicable to human eIF1A function and cancer. The eIF1A-NTT is highly conserved between two species, but the Kozak consensus and dependence on it are generally thought to vary between yeast and humans. The study, therefore, provides new insight into how the eIF1A-NTT maintains the fidelity of start codon selection, but the possible role of the cancer-associated residues in human eIF1A in mammalian start site selection is not determined. Ideally, this should be addressed directly, but at a minimum the end of the title should be altered to include \"in yeast\" to better describe the study.
Ribosome profiling of human cells altered by gene editing to express the R13P tumor mutation is being planned, but is beyond the scope of this paper. Hence, we have altered the title to include "in yeast".
> 2\) They must have become aware of a very relevant paper Johnson et al. PLoS One. 2017 Jun 8;12(6):e0178189). The authors did RNAseq of polysome fractions prepared from actual μm cancer cells harboring the eIF1A-NTT mutation. They identified genes whose translational efficiency was affected by the knockdown of EIF1AX. Could the authors look for relevant information regarding mRNAs from the list? The paper should be referenced and its implications discussed.
We have added new material to the Discussion to address these findings, stating: "A recent study on μm tumor cell lines expressing either WT or the G6D variant of EIF1AX provided evidence that the G6D substitution reduces the critical function of EIF1AX in bulk translation initiation. \[...\] Thus, it is possible that the reduction in RPG expression in G6D tumor cells is a response to reduced bulk translation and cell growth; and it will be interesting to determine whether the R13P substitution in EIF1X increases discrimination against AUGs in poor context in human cells."
> 3\) [Figure 5](#fig5){ref-type="fig"} convincingly demonstrates that the R13P and K16D mutations in eIF1A destabilize the closed/Pin conformation of the 48S PIC at UUG codons in vitro. Could they extend these results by showing a decreased UUG/AUG start codon usage in a cell-free translation system from eIF1A-mutant cells as compared to wildtype?
While we were not able to comply with this specific request, we were able to add pertinent new results that confirm the effects of both the R13P and K16D substitutions in reducing the UUG:AUG ratio in yeast cells using an independent assay based on luciferase reporters harboring UUG or AUG start codons. The new results are shown in the [Figure 3---figure supplement 1B](#fig3s1){ref-type="fig"}, and completely support the findings obtained using the *HIS4-lacZ* UUG and AUG reporters. Given the complete agreement between these two orthogonal in vivo assays and the effects of these substitutions in destabilizing the closed/Pin conformation of reconstituted 48S PIC at UUG codons in vitro, we hope the reviewers will agree that the evidence is very strong that these eIF1A substitutions increase discrimination against a near-cognate start codon.
| {
"pile_set_name": "PubMed Central"
} |
APR-246 (also called PRIMA-1^MET^) is the first compound in clinical development that reactivates mutant p53 in cancer cells by promoting its correct wild-type (wt) folding, thus triggering apoptosis.^[@bib1],\ [@bib2]^ The lead compound of APR-246, PRIMA-1, was originally discovered by Bykov *et al.*^[@bib3]^ APR-246 showed a good safety profile in a Phase I/II clinical dose-finding study on hematological malignancies and prostate cancer and both clinical and p53-dependent biological responses were observed.^[@bib4]^ A Phase Ib/II Proof of Concept study with APR-246 in combination with platinum-based therapy, in patients with recurrent p53-mutant high-grade serous (HGS) ovarian cancer, is ongoing. More than 96% of patients with HGS ovarian carcinoma carry TP53 mutations.^[@bib5]^
Platinum-based drugs have an important role in the treatment of many solid tumors including ovarian cancer. Cisplatin, the first drug of this class, has had a major impact in treatment of cancer but is also associated with severe adverse effects like nephrotoxicity. This prompted the development of the less toxic analog carboplatin.^[@bib6]^ The primary mechanism of action of platinum compounds is adduct formation with nucleophilic groups in tumor cell DNA. This triggers the DNA damage response pathway, in which p53 has a key role, leading to cell-cycle arrest, senescence and/or apoptosis.^[@bib7]^
Patients with ovarian cancer often respond well to the first-line platinum-based chemotherapy, but the majority of the patients with advanced stage tumors relapse and eventually die of chemotherapy-refractory disease. Platinum resistance is most often associated with decreased platinum levels at the site of action (i.e., DNA) and/or failure to trigger the DNA damage response after adduct formation.^[@bib6],\ [@bib7]^ The underlying molecular mechanisms of resistance to platinum compounds are multifactorial, involving drug-induced increase in cellular glutathione (GSH) levels leading to enhanced efflux of platinum compounds, reduced drug uptake, increased drug inactivation and DNA adduct repair, as well as inactivation of the tumor suppressor protein p53.^[@bib7],\ [@bib8],\ [@bib9],\ [@bib10]^ Mutation in p53 is one of the main mechanisms for inhibiting propagation of the DNA damage signal to the apoptotic machinery. About 50% of all tumors carry mutant p53 (see p53.free.fr, 2015) and cancer cells with defects in p53 are in general more resistant to conventional chemotherapy. In many tumors, including ovarian cancer, p53 mutations are correlated to shortened time to progression and decreased patient survival time.^[@bib11],\ [@bib12]^ Thus, restoration of wt function of p53 is a promising strategy for cancer therapy.^[@bib13],\ [@bib14]^
Here, we describe a new aspect of therapeutic activity of APR-246. APR-246 not only reactivates p53 but also decreases intracellular glutathione levels in a dose-dependent manner. Moreover, APR-246 completely restored cisplatin and doxorubicin sensitivity to mutant p53-carrying resistant ovarian cancer cells. Our results may open possibilities for greatly improved treatment of a wide range of platinum-resistant tumors.
Results
=======
APR-246 resensitizes cisplatin-resistant ovarian cancer cells to cisplatin
--------------------------------------------------------------------------
We first investigated whether APR-246 could resensitize the p53-mutant cisplatin-resistant A2780-CP20 and OVCAR-3 ovarian cancer cells to cisplatin using cell viability assay. The A2780-CP20 ovarian adenocarcinoma cell line carries a V172F mutation and was developed by chronic *in vitro* exposure of the parental A2780 cells to increasing concentrations of cisplatin.^[@bib15]^ The OVCAR-3 cells with hotspot p53 mutation (R248Q) were established from malignant ascites of a patient with progressive adenocarcinoma of the ovary.^[@bib16]^ The patient had been treated with cisplatin, doxorubicin and cyclophosphamide and was clinically resistant to cisplatin and doxorubicin.^[@bib16]^ Dose-response experiments with cisplatin alone and in combination with various concentrations of APR-246 were performed. As shown in [Figure 1a](#fig1){ref-type="fig"}, APR-246 resensitized A2780-CP20 cells to cisplatin in a dose-dependent manner. The IC~50~ value of cisplatin (with the partial effect contribution from APR-246 subtracted) decreased 18-fold from 52±11 to 3.2±0.8 *μ*M (mean±S.E.M.; *P*\<0.05; *t*-test), which is slightly lower than the IC~50~ value of cisplatin in A2780 cells (3.7±0.67 *μ*M). Thus, APR-246, at clinically relevant concentrations, completely restored the sensitivity of the ovarian cancer cells to cisplatin.
APR-246 also resensitized OVCAR-3 cancer cells to cisplatin ([Figure 1b](#fig1){ref-type="fig"}). The IC~50~ value of cisplatin decreased 3.2-fold, from 8.3±0.2 *μ*M to 2.6±0.9 *μ*M (mean±S.E.M.; *n*=2) in the presence of 20 *μ*M APR-246. Interestingly, in addition to increasing the sensitivity of the cells to cisplatin (i.e., decreasing the IC~50~ value), APR-246 appeared to increase the efficacy of cisplatin by reducing the survival index plateau at higher concentration from 30 to 5%.
Strong synergistic effects of APR-246 and platinum compounds in drug-resistant ovarian cancer cells
---------------------------------------------------------------------------------------------------
We then investigated whether APR-246 acts synergistically with cisplatin or carboplatin in cisplatin-resistant ovarian cancer cell lines. We found outstanding synergy (combination index (CI)\<0.3) with cisplatin ([Figure 2a](#fig2){ref-type="fig"}) or carboplatin ([Figure 2b](#fig2){ref-type="fig"}) in A2780-CP20 cancer cells. Outstanding synergy was also found in the mutant p53-carrying (Y126C and R337C) cisplatin-resistant ovarian cancer cell line IGROV-1/CDDP ([Supplementary Figure S1a and b](#sup1){ref-type="supplementary-material"}), which has been established by exposure of the parental wt p53-carrying IGROV-1 cells to cisplatin.^[@bib17]^ The IGROV-1 cell line was established from an untreated ovarian cancer patient.^[@bib18]^ Moreover, we observed strong synergistic effects with APR-246 and cisplatin in OVCAR-3 cells ([Figure 2c](#fig2){ref-type="fig"}). Furthermore, strong synergy (CI\<0.5) was found in the wt p53-carrying parental A2780 cell line, which was established from an untreated cancer patient^[@bib19]^ and outstanding synergy in the cisplatin-resistant A2780cis subline harboring wt p53 ([Supplementary Figure S1c and d](#sup1){ref-type="supplementary-material"}, respectively).^[@bib20]^ The results from these studies are summarized in [Table 1a](#tbl1a){ref-type="table"}. Finally, we investigated the effects of APR-246, cisplatin and their combination on apoptosis and reactive oxygen species (ROS) in OVCAR-3 cells. The synergistic response was evident based on emerging fractions of Annexin V+/PI− (early apoptotic) and Annexin V+/PI+ (late apoptotic/necrotic) cells ([Figure 2d](#fig2){ref-type="fig"}), as well as based on ROS induction ([Supplementary Figure S2](#sup1){ref-type="supplementary-material"}).
Cross-resistance
----------------
Treatment with cisplatin results not only in primary resistance but also in cross-resistance to other platinum compounds and classical alkylating agents, as well as anthracyclines including doxorubicin.^[@bib21]^ We performed dose-response experiments with cisplatin, carboplatin, doxorubicin and APR-246 in the A2780 line and its drug-resistant sublines A2780cis, A2780-CP20 and A2780ADR. The A2780ADR cells have wt p53 and have been developed by exposure of the A2780 cells to doxorubicin.^[@bib22]^ The results are summarized in [Table 1b](#tbl1b){ref-type="table"}. The IC~50~ values of cisplatin in A2780, A2780cis and A2780-CP20 were 3.7, 18 and 40 *μ*M, respectively. Thus, the IC~50~ value was increased 4.8-fold in the A2780cis cells carrying wt p53, and 11-fold in the mutant p53-carrying A2780-CP20 cells. The cisplatin-resistant sublines were cross-resistant to carboplatin and doxorubicin. The doxorubicin-resistant A2780ADR cells showed 18-fold resistance to doxorubicin and were cross-resistant to cisplatin and carboplatin. The IC~50~ value of APR-246 was less affected and was increased 1.6-fold in A2780-CP20 cells, whereas there was a 2-fold decrease in A2780ADR cells.
Synergistic effects of APR-246 and cisplatin in lung cancer cell lines
----------------------------------------------------------------------
We also tested the effect of APR-246 in combination with cisplatin in small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC) cell lines with various p53 mutations. Strong synergistic effects with APR-246 and cisplatin were observed in all cancer cell lines with hotspot p53 mutations (R248Q, R248W, R273H and G245C). These cells expressed high levels of mutant p53 ([Table 1a](#tbl1a){ref-type="table"}; [Supplementary Figure S3](#sup1){ref-type="supplementary-material"}). Mutant p53 often accumulates at high levels in cancer cells, which is believed to contribute to the strong apoptotic response upon APR-246 treatment.^[@bib3]^ Strong synergy was also seen in the SCLC cell line NCI-H378 with the Y163C p53 mutation that is not considered as a hotspot mutation but still occurs frequently in tumors. These cells expressed a lower level of p53 than the cells with hotspot p53 mutations ([Supplementary Figure S3](#sup1){ref-type="supplementary-material"}). Synergistic or strong synergistic effects were also observed in the lung cancer cell line NCI-H1417 with frameshift mutation in TP53 and no expression of full-length p53 ([Table 1a](#tbl1a){ref-type="table"}; [Supplementary Figure S3](#sup1){ref-type="supplementary-material"}).
Synergistic effects with APR-246 and cisplatin in primary ovarian cancer cells
------------------------------------------------------------------------------
Strong synergistic effects were observed in primary tumor cells from all five ovarian cancer patients included in the study ([Table 1c](#tbl1c){ref-type="table"}). DNA sequencing revealed that four of them had TP53 mutations. One of the mutations was the relatively frequently occurring Y220C mutation, whereas three patients had frameshift or nonsense mutations ([Table 1c](#tbl1c){ref-type="table"}).
*In vivo* antitumor effect of APR-246 in combination with cisplatin
-------------------------------------------------------------------
The antitumor effect of APR-246 in combination with cisplatin in mice bearing the aggressively growing A2780-CP20 tumor xenografts was examined. As shown in [Figure 3a](#fig3){ref-type="fig"}, single treatment with APR-246 and cisplatin inhibited tumor growth by 21 and 32%, respectively, while the combination resulted in 56% inhibition of tumor growth, indicating at least an additive effect. It should be noted that these doses were chosen to allow detection of a combination effect rather than to achieve maximal anticancer effect. Toxicity was evaluated on the basis of body weight reduction and observation of clinical signs of adverse effects. APR-246 was well tolerated and the general condition of the animals was good throughout the study. In the combination treatment group, the maximal body weight reduction was 10% and the mice recovered weight promptly after the treatment.
Using the same *in vivo* cancer model and treatment schedule, we examined the effect of combination treatment with APR-246 and cisplatin on activation of effector caspase-3, a marker of apoptosis. Analysis by immunohistochemistry showed an increase in active caspase-3-positive cells in all tumors ([Figure 3b](#fig3){ref-type="fig"}).
MQ is the active compound
-------------------------
APR-246 is a prodrug that is converted to MQ (2-methylenequinuclidin-3-one) and available evidence strongly suggests that MQ is the active compound responsible for the anticancer effects of APR-246.^[@bib1]^ To further investigate this, we compared the effect of MQ and APR-246 on cell viability of A2780-CP20 ovarian cancer cells. Both APR-246 and MQ reduced the A2780-CP20 cell viability in a dose-dependent manner ([Figure 4a](#fig4){ref-type="fig"}). MQ was 2.3-fold more potent than APR-246, with IC~50~ values of 4.8±0.4 *μ*M and 11±0.1 *μ*M, respectively (mean±S.E.M.; *P*\<0.05; *t*-test). In contrast, neither APR-320, a structural analog of APR-246 that cannot be converted to MQ, nor the MQ analog MQ-H that lacks Michael acceptor activity, had any effect on cell viability. Moreover, as shown in [Figure 4b](#fig4){ref-type="fig"}, MQ had strong synergistic effect with cisplatin in A2780-CP20 cells. These results are consistent with our previous results^[@bib1]^ and provide further support for MQ being the active compound.
APR-246 decreases intracellular glutathione levels
--------------------------------------------------
Many studies have shown that glutathione, which has an important role in maintaining the cellular oxidative balance, is involved in resistance to DNA-damaging drugs including platinum compounds and classical alkylating agents.^[@bib7],\ [@bib15],\ [@bib21],\ [@bib23],\ [@bib24],\ [@bib25]^ Intracellular glutathione exists in a balance between the reduced form (GSH), which constitutes the major fraction and is present at mM levels, and the oxidized form (GSSG). The drug-induced increase in intracellular glutathione concentration leads to, for example, increased efflux of cisplatin through ATP-binding cassette (ABC) transport pumps. A good correlation between intracellular glutathione levels and the degree of resistance to cisplatin has previously been shown in a panel of ovarian cancer cell lines, including the cell lines investigated in our study.^[@bib15]^
[Figure 5](#fig5){ref-type="fig"} shows that the glutathione levels were about 3-fold higher in the cisplatin-resistant A2780-CP20 cells than in the A2780 cells, in agreement with the previous results.^[@bib15]^ In both cell lines, APR-246 ([Figures 5a and b](#fig5){ref-type="fig"}) and MQ ([Figures 5c and d](#fig5){ref-type="fig"}) decreased glutathione in a dose-dependent manner, resulting in depletion of free glutathione at higher concentrations. Cisplatin alone did not have any significant effect on glutathione levels, while combination treatment with APR-246 and cisplatin resulted in more than additive effects ([Supplementary Figure S4](#sup1){ref-type="supplementary-material"}). Notably, APR-246 (50 *μ*M, 8 h) decreased glutathione levels equally (i.e., 2 nmol glutathione/10^6^ cells) in wild-type and p53-mutant cell lines ([Figures 5a and b](#fig5){ref-type="fig"}), suggesting that reactivation of p53 as such did not have any additional effect on glutathione levels.
MQ binds to glutathione
-----------------------
We then tested whether MQ reacts with glutathione. As shown in [Figures 5e and f](#fig5){ref-type="fig"}, MQ reacts rapidly with glutathione to form a Michael adduct. No reverse reaction was observed under these conditions. Analysis by HPLC-MS ([Figure 5e](#fig5){ref-type="fig"}) revealed a peak of the glutathione-MQ (GS-MQ) adduct at retention time 0.42, which was verified by the mass spectrometry (MS) ([Figure 5f](#fig5){ref-type="fig"}). No LC-MS signals corresponding to remaining MQ could be detected, indicating that the slight excess of GSH quickly consumed MQ.
Combination effects of APR-246 with doxorubicin
-----------------------------------------------
The main mechanisms of actions of anthracyclines, including doxorubicin, are inhibition of DNA and RNA synthesis by intercalation between base pairs of the DNA/RNA strands, and interference with the topoisomerase II enzyme, leading to double-strand breaks.^[@bib26]^ This results in a DNA damage response, including activation of the p53 pathway leading to apoptosis. Although it has been reported that mutation in p53 and the p53 pathway can cause resistance to anthracyclines,^[@bib27]^ the significance of p53 status for sensitivity to anthracyclines is less well documented than its role in the response to platinum compounds. We found that APR-246 completely restored the sensitivity of the A2780-CP20 cells to doxorubicin; the IC~50~ value of doxorubicin decreased 8-fold from 0.95±0.11 *μ*M to 0.12±0.01 *μ*M (mean±S.E.M.; *n*=2) ([Figure 6a](#fig6){ref-type="fig"}), which is equal to the IC~50~ value in the parental cells (0.12±0.05 *μ*M). APR-246 also resensitized OVCAR-3 cells to doxorubicin ([Figure 6b](#fig6){ref-type="fig"}). Similarly to the experiment with cisplatin ([Figure 1b](#fig1){ref-type="fig"}), APR-246 appeared to increase the efficacy of doxorubicin by reducing the survival index plateau. Moreover, we observed outstanding synergy with APR-246 in combination with doxorubicin in doxorubicin-resistant A2780ADR ovarian carcinoma cells ([Supplementary Figure S1f](#sup1){ref-type="supplementary-material"}). In the parental ovarian cancer A2780 cells, additive and synergistic effects with APR-246 and doxorubicin were observed ([Supplementary Figure S1e](#sup1){ref-type="supplementary-material"}).
Synergy with gemcitabine but not with docetaxel
-----------------------------------------------
The DNA-damaging drug gemcitabine is a nucleoside analog that replaces cytidine during DNA replication, leading to tumor growth arrest and eventually apoptosis. Strong synergistic effect with APR-246 and gemcitabine was observed in the A2780-CP20 cells ([Supplementary Figure S1g](#sup1){ref-type="supplementary-material"}). The mechanisms underlying the synergistic effect with APR-246 and gemcitabine have not been further investigated, but it has been reported that p53 and glutathione are involved also in resistance development for gemcitabine.^[@bib28]^
Notably, we did not observe any synergy between APR-246 and the taxane docetaxel in A2780-CP20 cells ([Supplementary Figure S1h](#sup1){ref-type="supplementary-material"}). Taxanes act by disrupting the function of the microtubules. Thus, they have a clearly different mechanism of action compared with the DNA-damaging drugs carboplatin, cisplatin, doxorubicin and gemcitabine. No clear role of p53 in the mechanism of action or resistance development to taxanes has been shown.^[@bib29],\ [@bib30]^
Discussion
==========
Platinum compounds are among the most effective anticancer drugs known, and have been used as a first-line treatment of several solid tumors, including ovarian cancer. Their main mode of action is interaction with DNA to form DNA adducts, which leads to a DNA damage response involving activation of p53-dependent apoptosis. However, repeated treatment with platinum drugs rapidly results in attenuation of the DNA damage response and resistance. Two of the main causes of resistance are p53 mutations and drug-induced increase in intracellular glutathione concentration.
The mode of action of APR-246 as a mutant p53-targeting anticancer compound is well documented.^[@bib1],\ [@bib2],\ [@bib31],\ [@bib32]^ APR-246 reactivates mutant p53 and induces expression of pro-apoptotic p53 target genes including Puma, Noxa and Bax, followed by activation of the mitochondrial apoptosis pathway.^[@bib32]^ APR-246 can also trigger apoptosis in a p53-independent manner by inducing ROS and endoplasmic reticulum (ER) stress^[@bib1],\ [@bib2]^ and by inhibiting thioredoxin reductase 1 (TrxR1) ([Figure 7](#fig7){ref-type="fig"}).^[@bib33]^ Recently, Tessoulin *et al.*^[@bib34]^ reported that APR-246 induced cell death in myeloma cells independently of p53 status by impairing the GSH/ROS balance.
APR-246 is a prodrug that is converted to the active compound MQ, a Michael acceptor and consequently a soft electrophile that reacts reversibly and preferentially with soft nucleophiles such as thiols in cysteines in p53. The p53 core domain has 10 cysteine residues to which MQ can potentially bind and stabilize p53 wild-type conformation.^[@bib1]^ Due to favorable molecular orbital interactions, high selectivity is achieved in comparison with classical hard electrophiles as the alkylating agents frequently used in cancer therapy. Certain other compounds that were identified based on their ability to target mutant p53-expressing cells are also Michael acceptors that can form adducts with thiol groups.^[@bib35],\ [@bib36]^ One of these, MIRA-1, had promising properties *in vitro* but is toxic *in vivo,* probably because it binds to multiple protein targets extracellularly, resulting in toxicity.^[@bib36]^ Thus, the optimal mutant p53-reactivating compound may be a prodrug such as APR-246 that is converted to the active compound intracellularly.
Our results show that MQ, in addition to binding to cysteines in p53, also binds to the cysteine in glutathione, a tripeptide formed by glutamic acid, cysteine and glycine, decreasing intracellular free glutathione levels in ovarian cancer cells. It is possible that MQ also binds to free cysteine and thereby inhibits glutathione synthesis. Moreover, APR-246/MQ induces formation of ROS in tumor cells,^[@bib1]^ which can lead to a further decrease in intracellular glutathione concentration. These multiple effects of APR-246/MQ presumably explain that APR-246, at clinically relevant concentrations, can deplete intracellular glutathione in ovarian cancer cells ([Figure 5](#fig5){ref-type="fig"}).
In addition to mutant p53, MQ can bind to unfolded inactive wt p53 and promote its correct folding.^[@bib1]^ It is conceivable that p53 protein unfolding leads to exposure of cysteine residues that can be modified. Indeed, APR-246 has been shown to activate wt p53 in melanoma cells in which p53 is inactivated by integrin αv-mediated signalling.^[@bib37]^ The proposed mechanisms of action can also explain the synergistic effects of APR-246 and platinum compounds observed in cisplatin-resistant cells that carry wt p53. In these cells, the synergy could be mainly due to decreased glutathione levels, although stabilization of wt p53 by MQ may contribute as well. In cancer cells that carry homozygous frame shift or nonsense mutations and therefore do not express full-length p53, the synergy could also be due to APR-246 effects on glutathione. Further studies are ongoing to further explore the molecular mechanism underlying the synergistic effects in cancer cells with various p53 status.
Most tumor-associated p53 mutations are missense mutations located in the DNA-binding core domain of p53. The most frequent p53 mutations, so-called hotspot mutations, affect amino-acid positions R175, G245, R248, R249, R273 and R282. In many cases, mutant p53 proteins have prolonged half-life and accumulate within cancer cells.^[@bib38]^ Many frequent mutations may also confer so-called gain-of-function activities to mutant p53.^[@bib38]^ We observed strong synergy with APR-246 and cisplatin in all cancer cells harboring homozygous hotspot mutations ([Table 1a](#tbl1a){ref-type="table"}). This is consistent with our previous studies showing that APR-246 can reactivate a wide range of mutant p53 proteins,^[@bib1],\ [@bib3],\ [@bib39]^ and is also consistent with data from us and others showing that PRIMA-1 and APR-246 can synergize with chemotherapeutic drugs including cisplatin and doxorubicin.^[@bib31],\ [@bib40],\ [@bib41]^ Here, we have explored combination treatment with APR-246 and DNA-damaging drugs in a more systematic and quantitative manner, using for example higher concentrations of APR-246 that result in stronger synergies. We have also used a broader range of cancer cell lines carrying different mutant forms of p53, with the aim of understanding the molecular mechanisms underlying the synergistic effects. Many of the cell lines examined in this study express high levels of mutant p53, which may contribute to the strong apoptosis-inducing effect of APR-246. Since p53 is a tetramer of four p53 monomers, heterozygous mutations may compromise the function despite the presence of a wt allele.^[@bib42]^
Resistance to chemotherapy is a major obstacle to clinical use of most chemotherapeutic agents, and numerous attempts to restore the chemosensitivity have been made. Much effort has focused on the ABC-drug transporters that have been shown to be overexpressed in many cancer cell lines as well as clinical samples, and cause reduced drug concentrations in cancer cells.^[@bib43]^ Co-administration of efflux pump inhibitors increases intracellular drug concentrations *in vitro*. However, clinical trials testing this paradigm have mostly failed,^[@bib44]^ presumably due to poor selectivity of the inhibitors resulting in intolerable side effects.^[@bib45]^ The glutathione system has also received attention, and based on encouraging results from animal models several clinical trials with buthionine sulfoximine (BSO), an inhibitor of glutathione synthesis, have been performed.^[@bib46]^ While BSO alone produced minimal toxic effects, combinations with melphalan occasionally resulted in severe myelosuppression.^[@bib47]^
Based on the results presented here we propose a unique mechanism of action of APR-246, in combination with platinum compounds, which distinguishes it from other anticancer drugs as well as other drugs that modulate platinum resistance ([Figure 7](#fig7){ref-type="fig"}). APR-246 is converted to MQ that binds to cysteine residues in mutant p53 and promotes refolding of the core domain. This provides a strong pro-apoptotic signal by itself, and also enhances the apoptotic response to platinum drugs that require functional p53 to exert their effect. In addition, MQ binds to the cysteine residue in glutathione and decreases intracellular glutathione concentration resulting in potentiation of the effect of platinum compounds. The synergistic effects observed with APR-246 and the DNA-damaging drugs doxorubicin or gemcitabine could at least in part be explained by the same mechanisms. This dual/multiple mechanism of action of APR-246 may provide a novel paradigm for overcoming platinum resistance in cancer therapy. Our results provide a strong rationale for the ongoing study with APR-246 in combination with carboplatin and pegylated liposomal doxorubicin in patients with recurrent ovarian cancer expressing mutant p53, and suggest that combination treatment with APR-246 and platinum or other DNA-damaging drugs could allow dramatically improved therapy of a wide range of therapy-refractory human tumors carrying mutant p53.
Materials and Methods
=====================
Test substances
---------------
APR-246 (2-hydroxymethyl-2-methoxymethyl-1-azabicyclo \[2,2,2\] octan-3-one) and MQ (2-methylenequinuclidin-3-one) were from Aprea (Solna, Sweden). Cisplatin (Ebewe or Hospira), carboplatin (Hospira), docetaxel (Actavis) and doxorubicin (Teva) were purchased from the Pharmacy at Akademiska sjukhuset, Uppsala, Sweden. Cisplatin was also purchased from Sigma (St. Louis, MO, USA). Gemcitabine was from LC Laboratories (Woburn, MA, USA), and glutathione from Sigma-Aldrich (Steinheim, Germany).
Cell lines and cell culturing
-----------------------------
Information about cell lines, including authentication and culture conditions, are described in [Supplementary Table S1](#sup1){ref-type="supplementary-material"}.
Primary cells
-------------
Human cancer tissue samples were obtained from Capital Biosciences (Rockville, MD, USA) and Genscript (Piscataway, NJ, USA). They had been enzymatically dispersed and filtered through 100--150 *μ*m filters, and the tumor cells were viable frozen and shipped to Aprea. The quality of the tumor cells was visually judged by a cytopathologist. According to quality criteria at least 70% of cells should be cancer cells and the cell viability should be at least 70%. Tissues were collected using Informed Consent, and the procedures were supported by ethics committees and were in accordance with the principles of the Declaration of Helsinki.
Analysis of TP53 gene status
----------------------------
All cell lines and primary samples were analyzed for *TP53* gene status (exons 2--11) by PCR amplification followed by single-strand conformation analysis (SSCA) according to the original protocol^[@bib48]^ and samples displaying gel mobility shifts were sequenced to confirm the nucleotide change. The analysis of *TP53* gene status was performed at Department of Clinical and Experimental Medicine, Linköping University.
Western blotting
----------------
Cells were analyzed by western blotting. Primary antibodies were anti-p53 antibody (\#9282, Cell Signaling, Danvers, MA, USA), anti-p53 antibody (\#FL-393, Santa Cruz, Dallas, TX, USA) and anti-GAPDH antibody (\#5632-1, clone EPR6256, Epitomics, Abcam, Cambridge, UK). Secondary antibodies were goat anti-mouse HRP-conjugated antibody (\#P 0447, Dako, Glostrup, Denmark), goat anti-rabbit HRP-conjugated antibody (\#P 044801-2, Dako). The experiments were performed at the Department of Clinical and Experimental Medicine, Linköping University.
Cell viability assays
---------------------
Cell viability assays used were FMCA, WST-1, Cell Titer-Glo and MTS assay. In all, 3000--12 000 cells/well in 96-well plates were incubated for 72 h, at 37 °C and with 5% CO~2~ before analysis. The FMCA and WST-1 assays were performed by Aprea, the Cell Titer-Glo assay by Accelera and the MTS assay by Oncodesign.
Annexin V/PI assay
------------------
OVCAR-3 cells were plated at a density of 75 000 cells per well in 3 ml of medium in 12-well plates. Next day, 2.5 ml medium was removed and cells were treated with cisplatin or APR-246 or in combination for 20 h. Next day, cells were harvested by trypsinization, washed twice and cells were stained with Annexin V and propidium iodine (PI) (FITC Annexin V apoptosis detection kit I, BD Biosciences, Stockholm, Sweden). After staining, the samples were analyzed by LSRII flow cytometer (BD Biosciences).
Measurement of intracellular ROS generation
-------------------------------------------
The procedure was done exactly as described in the experiment above until trypsinization, then cells were stained with 2′7′-dichlorofluorescin diacetate (DCF-DA) (Sigma) (5 μg/ml) in PBS with Ca^2+^/Mg^2+^ for 30 min at 37 °C. After staining the samples were analyzed by LSRII flow cytometer (BD Biosciences).
Analysis of results from combination studies
--------------------------------------------
For investigating possible additive and synergistic effects when using combinations of drugs, the data were analyzed with the Additive model.^[@bib49],\ [@bib50]^ Some of the results were also analyzed according to the Chou-Talalay model.^[@bib51]^ However, this model was less suitable for combination studies with APR-246 due to the considerably steeper dose-response curve with APR-246 than with platinum compounds and doxorubicin. In those studies where it could be used, similar results with Additive and Chou-Talalay model were obtained (results not shown). Factorial ANOVA model was used to evaluate the interaction between cisplatin and APR-246 in the apoptosis and ROS studies in OVCAR-3 cells.
*In vivo* xenograft efficacy study
----------------------------------
A2780-CP20 cells were injected s.c. into the left flank of female CD-1 Nu/Nu mice (5 × 10^6^ cells/mice) (Charles River, Italy). The mice were 5 weeks old and weighed 18--26 g. Treatment started when the mean tumor volume was \~120 mm^3^. APR-246 (in PBS) was administered as 2 h continuous i.v. infusion/day on treatment days and cisplatin (in water) as i.v. bolus injection immediately before APR-246 infusion on treatment days 2 and 6. Treatment volumes were 10 ml/kg. Tumors were measured with a caliper. Toxicity was evaluated on the basis of the body weight reduction. The mice were observed for clinical signs daily. The *in vivo* xenograft efficacy study was performed by Accelera. All procedures adopted for housing and handling of animals were in strict compliance with Italian and European guidelines for laboratory animal welfare.
Evaluation of active caspase-3 in tumors
----------------------------------------
Inoculation of tumor cells, strain and age of mice, handling procedures and ethical permissions were the same as in the *in vivo* efficacy study. Mouse weights were 20--24 g. Mice were randomized into two groups of three mice. To have homogeneity of tumor size, the mice were killed when the mean tumor size in each group was 0.68±0.25 cm^3^. At the scheduled time points, tumors were excised (in the treatment group 4 h after the end of the last infusion), formalin fixed, paraffin embedded, stained using primary monoclonal anti-active Caspase-3 antibody (Cell Signaling), and visualized by DAKO EnVision+Rabbit Polymer System and a Zeiss microscope (Axioscope-2 plus, Carl Zeiss, Göttingen, Germany). The stained tumor sections were examined in blind by two independent observers. The study was performed by Accelera.
Analysis of MQ binding to glutathione
-------------------------------------
GSH (97.7 *μ*mol) was added to a solution of MQ (73.0 *μ*mol) in 1 ml of de-ionized water containing NaHCO~3~ (\~20 mg) resulting in a pH of \~7. The mixture was stirred for 5 min at room temperature and analyzed by tandem HPLC-MS on an Agilent Series 1100 system using an ACE C8 (3 *μ*m, 3.0 × 50 mm) column and a mobile phase flow rate of 1 ml/min with a gradient of 10--97%, or 30--80% CH~3~CN in 10 mM NH~4~HCO~3~ buffer over 3 min. UV detection was performed at 220 nm. Electrospray mass spectrometry (ES-MS) was performed using an Agilent 1100 Series Liquid Chromatograph/Mass Selective Detector (MSD) to obtain the pseudo molecular \[M+H\]^+^ ion of the target molecules.
Glutathione assay
-----------------
In all, 30 000 cells/cm^2^ were seeded in 75 cm^2^ flasks (2.25 × 10^6^ cells/flask in 15 ml of medium) and incubated for 24 h before treatment with APR-246 for 4 and 8 h. Cells were harvested and counted, and intracellular total glutathione levels (glutathione \[GSH\] and glutathione disulfide \[GSSG\]) were measured using a Glutathione Assay Kit (Cayman Chemical Company, Ann Arbor, MI, USA). This kit contains glutathione reductase that converts GSSG to GSH. Cell diameters were measured using a cell Coulter counter, assuming spherical cells. No significant effect on cell viability was observed at the indicated concentrations of APR-246 and MQ. These experiments were performed by Accelera. For more detailed information about the Materials and Methods, see [Supplementary Methods](#sup1){ref-type="supplementary-material"}, additional Information.
We thank Drs. Mikael von Euler (Aprea), Ninus Caram-Lelham (Aprea), Thierry Soussi (Karolinska Institutet), Rolf Larsson (Uppsala University), Peter Söderkvist (Linköping University) and Enrico Pesenti (Accelera) for helpful discussions.
ABC
: ATP-binding cassette
APR-246
: 2-hydroxymethyl-2-methoxymethyl-1-azabicyclo \[2,2,2\] octan-3-one
BSO
: buthionine sulfoximine
CI
: combination index
GSH
: glutathione
GSSG
: glutathione disulfide
HGS
: high-grade serous
i.v.
: intravenous
MS
: mass spectrum
MQ
: 2-methylenequinuclidin-3-one
NSCLC
: non-small cell lung cancer
PI
: propidium iodine
ROS
: reactive oxygen species
s.c.
: subcutaneous
SCLC
: small cell lung cancer
SSCA
: single-strand conformation analysis
TrxR1
: thioredoxin reductase 1
Wt
: wild type
[Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies this paper on Cell Death and Disease website (http://www.nature.com/cddis)
Edited by Z-X Xiao
NM, JA, ÅF, MU and UB are employed or have been employed at Aprea. KGW is a co-owner and board member of Aprea. VJNB is a co-owner of Aprea. SB is a consultant at Aprea.
Supplementary Material {#sup1}
======================
######
Click here for additional data file.
{#fig1}
{#fig2}
{#fig3}
{#fig4}
![APR-246 and MQ reduce glutathione levels in ovarian cancer cells. (**a**--**d**) Effects of APR-246 and MQ on intracellular glutathione levels in A2780 and in A2780-CP20 ovarian cancer cells. Total GSH levels (GSH+2 × GSSG) were measured using Cayman Glutathione Assay kit. The results are shown as mean±S.E.M. (*n*=2). (**e** and **f**) MQ forms a Michael adduct with glutathione. (**e**) Liquid chromatography trace of the reaction mixture after addition of excess glutathione (GSH) shows total consumption of MQ. Product and GSH co-eluates under the conditions used. (**f**) Mass spectrometry (MS) of the retention time 0.42 peak shows the typical MS pattern of the GS-MQ adduct (m/z: 445 +223). MS-peak of GSH (in excess): m/z: 308 \[m+H\]. The typical MS pattern of the GS-MQ adduct (m/z: 445 and 223) and the slight excess of added GSH to the reaction mixture is shown by the MS peak m/z=308 \[m+H\]. The results shown in (**e**--**f**) are representative of three independent experiments](cddis2015143f5){#fig5}
{#fig6}
{#fig7}
###### Results from combination studies with APR-246 and cisplatin in ovarian and lung cancer cell lines
**Cancer cell lines** **Cancer type** **p53 status** **p53 protein expression** **Combination APR-246 and cisplatin**
----------------------- ----------------- --------------------------------- ---------------------------- ---------------------------------------
*Ovarian*
OVCAR-3 Ovarian R248Q (hom.) +++ SS
A2780-CP20 Ovarian V172F (het.) \+ SS
IGROV-1/CDDP Ovarian Y126C (het.), R337C (het.) \+ S/SS
A2780 Ovarian wt − S/SS
A2780cis Ovarian wt − SS
A2780ADR Ovarian wt − Add/S
*Lung*
NCI-H1770 NSCLC R248W (hom.) +++ SS
NCI-H1975 NSCLC R273H (hom.) +++ SS
NCI-H596 NSCLC G245C (hom.) +++ SS
NCI-H378 SCLC Y163C (hom.) \+ SS
NCI-H1417 SCLC 175fs246\* (hom.) ('p53 null\') − S/SS
Abbreviations: hom., homozygous; het., heterozygous; \*, stop codon; fs, frame shift; 'p53 null\', no full-length p53; −, no p53 expression seen; +, weak p53 expression; +++, strong p53 expression; SCLC, small cell lung cancer; NSCLC, non-small cell lung cancer; Add, additive (CI=1.0±0.2); S, synergy (CI\<0.8); SS, strong synergy (CI\<0.5).
MTS, FMCA or Cell Titer-Glo assays were used for measurement of cell viability. CI was calculated using Additive model. It should be noted that the sequencing method used (Sanger sequencing and Single Strand Conformation Analysis) cannot distinguish between homozygous and hemizygous mutations. Also, 'het.\' refers to that both wt p53 and mut p53 are found in the sample. This can either be due to heterozygosity or to a presence of cells with different p53 status, which is not common in cancer cell lines but may occur in primary cancer cells
###### IC~50~ values and resistance factors for APR-246, platinum compounds and doxorubicin in the parental ovarian A2780 cell line and drug-resistant sublines
**Substance** **A2780 (wt p53) IC**~**50**~ **(μM)** **A2780cis (wt p53) IC**~**50**~ **(μM)** **Resistance factor** **A2780-CP20 (het. V172F p53) IC**~**50**~ **(μM)** **Resistance factor** **A2780ADR (wt p53) IC**~**50**~ **(μM)** **Resistance factor**
--------------- ---------------------------------------- ------------------------------------------- ----------------------- ----------------------------------------------------- ----------------------- ------------------------------------------- -----------------------
Cisplatin 3.7±0.66 18±1.9\*\*\* 4.8 40±4.6\*\* 11 15±1.5\*\* 4.2
Carboplatin 76±13 170±10\*\*\* 2.2 425±42\*\*\* 5.6 17±16\*\* 2.3
Doxorubicin 0.12±0.047 0.31±0.054\* 2.6 0.76±0.060\*\*\* 6.4 2.1±0.59\* 18
APR-246 23±1.6 20±1.1 0.83 37±2.3\*\*\* 1.6 11±2.0\*\*\* 0.48
FMCA was used for measurement of cell viability. *t*-test (two tailed, unpaired, unequal variance) was used for statistical analysis of differences in potency (IC~50~ values) of drugs in drug-resistant sublines compared with the parental cell line A2780; \**P*\<0.05; \*\**P*\<0.01; \*\*\**P*\<0.001. The results are mean±S.E.M. of at least three independent experiments
###### Results from combination studies with APR-246 and cisplatin in primary ovarian cancer cells
**Patient number** **Histological description** **p53 status** **Combination APR-246 and cisplatin**
-------------------- -------------------------------------- ------------------------------------- ---------------------------------------
1 Serous adenocarcinoma, grade 2 P153H fs 180\* (hom.) ('p53 null\') SS
2 Serous adenocarcinoma, grade 3 C135A fs 169\* (het.) SS
3 Serous adenocarcinoma Y220C (hom.) SS
4 Poorly differentiated adenocarcinoma wt p53 SS
5 Adenocarcinoma Q165\* (het.) SS
Abbreviations: hom., homozygous; het., heterozygous; \*, stop codon; fs, frame shift; 'p53 null\', no full-length p53; SS, strong synergy (CI\<0.5).
FMCA assay was used for measurement of cell viability. CI was calculated using Additive model. It should be noted that the sequencing method used (Sanger sequencing and Single Strand Conformation Analysis) cannot distinguish between homozygous and hemizygous mutations. Also, 'het.\' refers to that both wt p53 and mut p53 are found in the sample. This can either be due to heterozygosity or to a presence of cells with different p53 status, which is not common in cancer cell lines but may occur in primary cancer cells
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Hidrotic ectodermal dysplasia (HED), also known as Clouston syndrome, is a rare autosomal dominant genetic skin disorder ([@B01]). It is characterized by alopecia, nail dystrophy, and palmoplantar hyperkeratosis ([@B02]). Mutations in the GJB6 gene cause HED, hereditary autosomal dominant non-syndromic deafness, and keratitis-ichthyosis-deafness syndrome. This gene encodes the gap junction protein connexin 30 (Cx30) ([@B03]). Cx30 is a member of the large gap-junction protein family, and it plays a role in the homeostasis of the epidermis and inner ear through gap junction intercellular communication. Lamartine et al. first confirmed that *GJB6* is the disease-causing gene of HED ([@B04] [@B05]). To date, five mutations have been found in patients with HED: *G11R*, *V37E*, *D50N*, *A88V*, and *N14S* in *GJB6* ([@B03]--[@B06]). We previously reported the *G11R* mutation in a Chinese family that caused HED of the hair and nails only ([@B07]).
In the process of our experiments, we found that the expression of *GJB6* mutants induced HatCaT cell death within 48 h ([@B08]). Berger et al. ([@B09]) determined that the *A88V* mutant caused cell death through membrane disruption within 24 h of expression. In this study, we investigated the mechanistic effects of the human *GJB6* gene and its mutants by constructing lentiviral vectors containing human wild-type *GJB6* and the variants *A88V* and *G11R*. We then stably expressed these proteins via the Tet-on system in the HaCaT cell line ([@B10]). This facilitated research on the mechanisms of apoptosis induction by the expression of these mutants in HaCaT cells.
Material and Methods
====================
Cell culture
------------
The HaCaT cell line and the Tet-on expression system were used to stably express vectors containing the wild-type *GJB6* gene and its mutant variants *A88V* and *G11R*. The vector inserts were obtained by whole gene synthesis. The sequences used were GJB6: NM_006783 (A88V) and GJB6: NM_006783 (G11R). The cells were cultured in RPMI 1640 medium (Gibco, USA) supplemented with 10% fetal bovine serum (Gibco) in a humidified incubator maintained at 37°C under 5% CO~2~. Once the cells had reached 80--100% confluence, the cells were detached using 0.25% trypsin-EDTA (Hyclone, USA) and then passaged.
Cell proliferation
------------------
Cell proliferation was measured with the CCK8 assay (Dojindo, China). The HaCaT cells, which were transfected with either the negative control (NC) virus, the wild-type *GJB6* gene (WT), or the *A88V* and *G11R* mutants, were divided into two groups: 1) control group: NC, WT, A88V, and G11R cultured without doxycycline (free DOX); and 2) experimental group: NC, WT, A88V, and G11R cultured with DOX. The HaCaT cells were aliquoted into 96-well plates and incubated overnight to allow cell attachment. After 4, 8, 12, 24, 36, and 48 h of incubation following the addition of DOX or free DOX, the cells in each well were incubated with 10 μL of CCK8 for 1 h to measure the levels of cell proliferation.
Flow cytometry analysis
-----------------------
FITC annexin V and propidium iodide (PI) staining were performed using the FITC annexin V Apoptosis Detection Kit from BD Biosciences (USA) to determine the level of apoptosis. The control and experimental group were first induced with DOX and free DOX, respectively. After 8 h, the cells were collected from the plates by trypsinization and washed twice with PBS. The cells were then resuspended in 100 μL binding buffer and 5 μL FITC annexin V and 5 μL PI were added to each well as per the manufacturer\'s instructions. After 15 min incubation in the dark, 400 μL binding buffer was added to the cells and the FITC annexin V/PI-stained cells were analyzed by flow cytometry. The percentages of apoptotic cells were measured with the BD FACSDiva 7.0 flow cytometer (BD Biosciences).
Western blot analysis
---------------------
Western blot analyses were performed to analyze the sequence-specific effect of Cx30 on the expression of apoptosis-related proteins in the transfected HaCaT cells. Briefly, 12 h post treatment, cells were lyzed in RIPA buffer (Solarbio, China), and proteins were separated by SDS-PAGE and transferred to microporous polyvinylidene difluoride (PVDF) membranes (Solarbio). The membranes were incubated in 5% Blotto non-fat dry milk (Santa Cruz Biotechnology, USA) with 0.05% Tween-20 (Sigma-Aldrich, USA) in PBS (PBS-T) for 1 h at room temperature and subsequently incubated overnight at 4°C with primary antibodies against Cx30 (1:1000, rabbit, Invitrogen, USA), BAX (1:1000, rabbit, Cell Signaling Technology, USA), caspase-3, cleaved caspase-3 (1:1000, rabbit, Cell Signaling Technology), and β-actin (1:1000, mouse, Sigma-Aldrich). Before incubation with HRP-conjugated goat anti-mouse or anti-rabbit secondary antibodies, the membranes were washed three times (10 min each) in PBS-T buffer. The blots were then scanned and densitometry measurements were carried out with the LAS 4000 gel-imaging system (Fujifilm, Japan). Each signal was normalized to the β-actin loading control in the same lane.
Statistical analysis
--------------------
Data are reported as means±SD. Statistical significance was defined as P\<0.05 using Student\'s *t*-test.
Results
=======
Changes in cell morphology after induction with doxycycline (DOX)
-----------------------------------------------------------------
Cellular morphological changes were observed under the microscope (100×) after induction by DOX. The morphology of cells transfected with the negative control virus or wild-type *GJB6* remained unchanged after the cells were induced. However, the morphology of cells expressing the *A88V* and *G11R* mutants became senescent and the cell nuclei underwent pyknosis. The cell morphology changes were exhibited by the mutant-expressing cells within 8 h of DOX induction and the cells were mostly dead after 48 h. The surviving cells after 48 h were senescent with pyknotic cell nuclei ([Figure 1](#f01){ref-type="fig"}).
{#f01}
Reduction of HaCaT cell viability in cells transfected with the *A88V* and *G11R* mutants
-----------------------------------------------------------------------------------------
We measured the absorbance of each group of cultured HaCaT cells with the CCK8 assay after 4, 8, 12, 24, 36, and 48 h incubation after induction ([Table 1](#t01){ref-type="table"}). The results indicated that the cell viability of HaCaT cells transfected with the negative control virus (NC) and induced by DOX were not significantly different from the NC cells that were not induced by DOX (P\>0.05) ([Figure 2A](#f02){ref-type="fig"}). Cells expressing the wild-type *GJB6* gene induced by DOX after 4, 8, 12, 24, and 36 h incubation were significantly more viable than the same cells not induced by DOX (P\<0.05) ([Figure 2B](#f02){ref-type="fig"}). In contrast, the DOX-induced cells expressing the *A88V* and *G11R* mutants after 4, 8, 12, 24, 36, and 48 h incubation were significantly less viable than the same cells not induced by DOX (P\<0.05) ([Figure 2C and 2D](#f02){ref-type="fig"}). Furthermore, the results of the LSD test showed that the DOX-induced cells expressing the *A88V* and *G11R* mutants were significantly less viable compared with the cells expressing the wild-type *GJB6* gene or the negative control virus across all time-points ([Figure 2E](#f02){ref-type="fig"}). The CCK8-based cell proliferation assay showed no significant difference in proliferation between the G111R- and A88V-expressing cells, while these two groups of cells did show significantly reduced cell proliferation in comparison with the NC and wild-type groups.
{#f02}
Table 1.Cell growth inhibition evaluated by CCK8 analysis of HaCat cells induced or not by doxycycline (+DOX, and -DOX).Group4 h8 h12 h24 h36 h48 hNC+DOX0.469±0.0350.572±0.1100.645±0.0320.916±0.0361.390±0.2131.878±0.357-DOX0.492±0.0180.596±0.0700.679±0.1490.957±0.0931.370±0.2381.850±0.287t1.6830.5540.6671.2490.1840.180P0.1120.5870.5140.2300.8560.860WT+DOX0.281±0.0110.336±0.0270.340±0.0430.494±0.0250.701±0.0521.088±0.198-DOX0.321±0.0140.395±0.0220.438±0.0650.595±0.0111.118±0.1861.256±0.213t6.6305.1414.64410.6656.4621.729P**0.0000.0000.0000.0000.000**0.103*A88V*+DOX0.310±0.0070.221±0.0210.160±0.0140.136±0.0020.139±0.0040.148±0.009-DOX0.368±0.0290.616±0.0810.489±0.0420.694±0.0341.124±0.0971.690±0.172t5.80914.23122.14248.19830.44426.845P**0.0000.0000.0000.0000.0000.000***G11R*+DOX0.250±0.0130.222±0.0210.181±0.0180.142±0.0020.134±0.0030.123±0.003-DOX0.306±0.0170.433±0.0310.438±0.0530.669±0.0381.084±0.0901.454±0.220t7.76917.06413.87541.82731.78118.127P**0.0000.0000.0000.0000.0000.000**[^2]
Apoptosis induction by the *A88V* and *G11R* mutations of the *GJB6* gene
-------------------------------------------------------------------------
To determine whether the growth inhibitory effect of the *A88V* and *G11R* mutants of the *GJB6* gene is associated with apoptosis, annexin V/PI staining of HaCaT cells expressing the wild-type *GJB6* gene and the *A88V* and *G11R* mutants were analyzed with flow cytometry ([Figure 3](#f03){ref-type="fig"}). The cells were classified as Q2 necrotic cells (annexin V+, PI+), Q3 healthy cells (annexin V−, PI−) and Q4 apoptotic cells (annexin V+, PI−). The HaCaT cells expressing the *A88V* and *G11R* mutants showed increased apoptosis and lower cell viability that was significantly different (P\<0.05) than the cells transfected with the negative control virus or the wild-type *GJB6* gene.
{#f03}
Induction of protein cleavage by the *A88V* and *G11R* mutants
--------------------------------------------------------------
Expression of the *A88V* and *G11R* mutants resulted in elevated levels of the cleaved forms of caspase-3, caspase-8, caspase-9, and PARP, which are all markers of apoptotic activation while the same markers were unchanged in cells transfected with the negative control virus or the wild-type *GJB6* gene. The level of BAX remained unchanged in all groups.
[Figure 4](#f04){ref-type="fig"} shows the effects of the *A88V* and *G11R* mutations of *GJB6* on apoptosis-related proteins.
{#f04}
Discussion
==========
In the present study, we discovered that the *A88V* and *G11R* mutants induced apoptosis of HaCaT cells, which quickly led to population-wide cell death. To further research the mechanistic effect of the human *GJB6* gene and mutations on the proliferation and apoptosis of HaCaT cells, we constructed lentiviral vectors containing the human *GJB6* gene and the mutants, and stably expressed these proteins using the Tet-on system in the HaCaT cell line ([@B10]). The Tet-on system has potential for numerous applications, including the study of gene function and gene therapy. This system enables the efficient expression of genes following induction by DOX ([@B11]). It also mitigates the poor regulation of gene expression of the many other expression systems, as our team experienced when we first constructed the traditional overexpression lentivirus vector ([@B08]).
Apoptosis is a conserved phenomenon that plays a critical role in the regulation of the cellular activities of eukaryotes, and the process is characterized by chromatin condensation ([@B12],[@B13]). In our study, the HaCaT cells expressing the *A88V* and *G11R* mutants were found to exhibit significantly increased apoptosis. Moreover, the cells stained with annexin V and PI indicate apoptotic cells (annexin V) and dead or late apoptotic cells (PI). Early apoptotic cells are PI-negative and annexin V-positive, while dead or apoptotic cells are positive for both annexin V and PI. Based on the cell staining results, the *A88V* and *G11R* mutants clearly increased apoptosis in HaCaT cells.
There are two major apoptotic pathways, classified as extrinsic or intrinsic cell death ([@B13],[@B14]). The results of the present study indicated that the *A88V* and *G11R* mutants induced the activation of caspase-3, caspase-8, caspase-9, and PARP. However, there was no change in BAX activity compared with the control indicating that this protein may play a minor role at most in the apoptosis of HaCaT cells ([@B15]). Much evidence suggests that the apoptotic process is triggered by the activation of caspases in various cell types ([@B16]).
Caspase, a cysteinyl aspartate-specific proteinase, is an important player in cell apoptosis. Caspases are divided into two broad categories: the initiator caspases and the effector caspases ([@B17],[@B18]). Apoptotic initiators (such as caspase-2, -8, -9, and -10) activate the downstream cascade of necroptosis with the help of the adapter proteins. Active caspase-8 then either initiates apoptosis directly by cleaving and thereby activating executioner caspases, or by activating the intrinsic apoptotic pathway through the cleavage of BID to induce efficient cell death ([@B19],[@B20]). Cascade-initiating pathways converge on the activation of the downstream effector caspases (such as caspase-3, -6, and -7) that act to kill the cells by cleaving death substrates.
As previously reported, caspase-3 is a member of the activated-caspase family ([@B21]). It is the major activated factor in the process of apoptosis, and its activation is a sign that apoptosis has entered the irreversible phase. This protein is therefore considered the end executor of apoptosis. Usually, the extrinsic apoptosis pathway is activated through the binding of a ligand to a death receptor, which is mediated by caspase-8, while the intrinsic apoptotic pathway is mediated by mitochondrial cytochrome c and caspase-9. Caspase is synthesized as an inactive proenzyme (zymogen) in living cells. The proenzyme is activated through cleavage, and this active form is involved in nuclear condensation, cellular detachment, and phosphatidylserine externalization, all hallmark characteristics of apoptosis. These effector caspases then cleave PARP ([@B12]), and the activated PARP confers apoptotic cells with their classic morphological and biochemical characteristics.
In summary, our results showed that the *A88V* and *G11R* mutants of *GJB6* may activate the downstream execution factor, caspase-3, both through the extrinsic apoptotic pathway mediated by caspase-8 and the intrinsic apoptotic pathway mediated by caspase-9. This rapidly induced HaCaT cell apoptosis by means of the lyase, PARP. Further investigations are required to determine whether the *A88V* and *G11R* variants can also induce apoptosis through pathways other than those mediated by caspases, and to determine how these mutants lead to the clinical manifestations of HED.
This research was supported by the following grants: National Natural Science Foundation of China (grant number 81171492/H1102), The Natural Science Foundation of Shandong Province for Youths (grant number ZR2011HQ056), and The Science and Technology Development Program of Shandong Province (grant numbers 2010GSF10812 and 2011GSF11847). The central laboratory of Shandong Provincial Hospital is affiliated to Shandong University. We thank Natasha Beeton-Kempen, from Liwen Bianji, Edanz Editing China (www.liwenbianji.cn/ac), for editing the English text of a draft of this manuscript.
[^1]: \*These authors contributed equally to this study.
[^2]: Data are reported as means±SD absorbance values for n=9. NC: Negative control; WT: wild type; A88V: *A88V* mutant; G11R: *G11R* mutant. Statistical analysis was done with the *t*-test.
| {
"pile_set_name": "PubMed Central"
} |
Since the rosiglitazone controversy aroused in 2007 \[[@B1]\], US Food and Drug Administration and most other regulatory agencies have requested to perform cardiovascular safety assessments for upcoming antidiabetic medications \[[@B2]\]. Dipeptidyl peptidase-4 (DPP4) inhibitors belong to a new class of oral antidiabetic drugs which can ameliorate hyperglycemia by increasing endogenous concentration of glucagon-like peptide-1, a gut-derived hormone that stimulates insulin secretion and delays gastric emptying. However, despite its popularity in the diabetes market, the cardiovascular safety of DPP4 inhibitors has not been investigated following the two large, randomized, placebo-controlled trials (RCTs), saxagliptin assessment of vascular outcomes recorded in patients with diabetes mellitus-thrombolysis in myocardial infarction (SAVOR-TIMI 53) \[[@B3]\] with saxagliptin and Examination of Cardiovascular Outcomes with Alogliptin vs. Standard of Care (EXAMINE) \[[@B4]\] with alogliptin, which recently published the results of cardiovascular outcomes in 2013.
Unexpectedly, SAVOR-TIMI 53 showed that the rate of hospital admission for heart failure was significantly increased by 27% in subjects treated with saxagliptin, while EXAMINE did not report any outcome related to heart failure. A subsequent paper based on EXAMINE trial reported that alogliptin was not associated with the increased risk of heart failure outcomes, such as hospital admission for heart failure, compared with placebo group \[[@B5]\]. However, both studies observed that hospital admission for heart failure occurred more frequently in subjects treated with saxagliptin or alogliptin who had no history of this disease condition (SAVOR-TIMI 53: hazard ratio \[HR\], 1.30, 95% confidence interval \[CI\], 1.03 to 1.65; EXAMINE: HR, 1.76, 95% CI, 1.07 to 2.90). According to the additional report from SAVOR-TIMI 53, the high-risk patients for heart failure hospitalization were those who had a prior history of heart failure, elevated baseline levels of N-terminal pro-brain natriuretic peptide, or chronic kidney disease \[[@B6]\].
To date, several studies demonstrated conflicting results regarding the safety of DPP4 inhibitors on heart failure. A nested case-control study based on a large insurance claimed that the database from United States showed that sitagliptin was associated with heart failure admissions in patients recently diagnosed with heart failure (adjusted odds ratio \[OR\], 1.84; 95% CI, 1.16 to 2.92; *P*=0.01) \[[@B7]\]. A propensity-matched analysis using Taiwan National Health Insurance research database with a total of 16,576 subjects reported that sitagliptin users were hospitalized for heart failure more frequently than patients that were never exposed to a DPP4 inhibitor (HR, 1.21; 95% CI, 1.04 to 1.42; *P*=0.017) \[[@B8]\]. In this study, patients with the highest adherence to sitagliptin had the greatest risk of developing heart failure (HR, 2.56; 95% CI, 2.10 to 3.12) compared with never users. In addition, similar findings were derived from a recent meta-analysis of 82 randomized controlled clinical trials with various DPP4 inhibitors treated for at least 24 weeks. The overall risk of acute heart failure was significantly elevated in the DPP4 inhibitor group compared with placebo/active comparators (*n*=69,615; Mantel-Haenszel OR, 1.19; 95% CI, 1.03 to 1.37; *P*=0.015) \[[@B9]\]. In Vildagliptin in Ventricular Dysfunction Diabetes (VIVIDD) trial, 254 subjects with type 2 diabetes and New York Heart Association functional classes I through III heart failure were randomized to vildagliptin or placebo for 12 months \[[@B10]\]. Although there was no difference in left ventricular ejection fraction between two groups, patients treated with vildagliptin had an increase in left ventricular end-diastolic volumes, suggesting that vildagliptin could possibly increase the risk of heart failure.
However, a recent population-based study from United Kingdom showed that incretin-based drugs were not associated with an increased risk of heart failure among patients with type 2 diabetes (OR, 0.85; 95% CI, 0.62 to 1.16) \[[@B11]\]. Although five drugs (sitagliptin, vildagliptin, saxagliptin, exenatide, and liraglutide) were included in the analysis, majority of patients were prescribed sitagliptin. A non-randomized, observational real-world study from Germany (DiaRegis registry) enrolled patients with prior metformin monotherapy who then received either DPP4 inhibitors or sulfonylurea \[[@B12]\]. Similarly, there was no difference in event rates of heart failure between two groups (1.7% in the DPP4 inhibitor group vs. 2.0% in the sulfonylurea group, *P*=0.85), while subjects with DPP4 inhibitors showed significantly lower rates of stroke and transitory ischemic attack.
In this issue of *Diabetes and Metabolism Journal*, Seo et al. \[[@B13]\] provided additional concerns about the safety of DPP4 inhibitors on heart failure in Korean type 2 diabetes. Based on Korean Health Insurance claims database, they used pioglitazone as a positive control for increasing risk of heart failure \[[@B14]\] and compared the incidence rate of hospital admission for heart failure with DPP4 inhibitors. Compared with pioglitazone group, sitagliptin and vildagliptin users did not show any significant difference in hospitalization rate for heart failure, indicating the comparable effects on the development of heart failure between DPP4 inhibitors and pioglitazone. Despite the large number of subjects, the potential for confounding in this study cannot be ruled out because of the limitation of the nature of claim database. Furthermore, there is no information regarding compliance of medications and comorbidities in the study population which should be adjusted in the analysis. Regardless of drug types, the hospitalization rate for heart failure was greatest in the first 30 days after starting to take the medication, when compared with days 31 to 360 days. Similar to their findings, Scirica et al. \[[@B6]\] reported that the risk of heart failure hospitalization was most pronounced in the first 12 months of treatment with saxagliptin. These two findings suggest that DPP4 inhibitors may trigger heart failure in vulnerable but unidentified or subclinical subjects who are likely to develop heart failure. On the contrary to thiazolidinediones, DPP4 inhibitors are not associated with edema or fluid retention, therefore the mechanism for heart failure might be different, which should be elucidated by further research.
In summary, inconsistent findings from various studies, including meta-analyses, large randomized trials, observational studies, and analysis of insurance claim database, indicate that we still cannot draw any definite conclusion in terms of the safety of DPP4 inhibitors on heart failure in patients with type 2 diabetes. We should wait with prudence for the upcoming reports from Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS) trial \[[@B15]\], another large RCT with Sitagliptin including 14,724 subjects with type 2 diabetes and Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results (LEADER) trial \[[@B16]\], to fully address all the concerns related to the safety issues of incretin-based drugs on heart failure.
**CONFLICTS OF INTEREST:** No potential conflict of interest relevant to this article was reported.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Bone Targeting Agents (BTAs), such as bisphosphonates and denosumab, can reduce the risk of skeletal-related adverse events and are widely used to protect the skeleton in patients with osteoporosis or bone metastases from various types of cancer \[[@ref1], [@ref2]\]. Osteonecrosis of the jaw (ONJ), a rare but potentially serious adverse event, has been associated with cumulative doses of BTAs and may result in BTAs interruption. More than 90% of cases of ONJ occur in cancer patients, who receive high doses of i.v. bisphopshonates or denosumab. Appropriate prevention measures can reduce the risk of ONJ, while early diagnosis can lead to successful management.
ONJ and ecancer e-learning module
=================================
Since the first reports of ONJ, in 2003, an increasing number of research data has been published and dedicated researchers continuously produce new knowledge. Osteonecrosis is classically an area of painful exposed necrotic bone on the mandible or maxilla. Today, we know that ONJ may also present as non-exposed bone \[[@ref3], [@ref4]\]. Cases of ONJ associated with anticancer drugs, which include classical chemotherapy agents, inhibitors of angiogenesis, tyrosine kinase and mammalian target of rapamycin inhibitors and immunotherapy agents have been reported \[[@ref5]\]. The pathogenesis of ONJ in most likely multifactorial and infection seems to play an important role. Critical questions on ONJ have been clarified, while recently described modified surgical protocols seem to reduce the risk of ONJ following dental extraction \[[@ref6], [@ref7]\]. Only 3 among 89 patients with cancer receiving antiresorptives developed osteonecrosis following dental extractions, which were performed with alveolectomy and primary surgical closure \[[@ref6]\]. Localized dental and periodontal disease, with histological necrosis of the alveolar bone may precede the clinical appearance of necrotic bone \[[@ref8], [@ref9]\]. This knowledge has re-enhanced our ONJ prevention protocols for oral health maintenance and has changed the globally accepted previous instruction for the avoidance of dental extraction in patients, who receive BTAs. ONJ may develop prior to dental extraction, while dental and/or jaw pain, swelling and tooth mobility are associated signs and symptoms \[[Figure 1](#figure1){ref-type="fig"}\]. As it is obvious, early ONJ, prior to dental extraction, may mimic localized dental or/and periodontal disease and infection and thus, it may be difficult to diagnose. Clinicians need adequate knowledge, obtained with education, in order to have a high suspicion index for ONJ early diagnosis. Early diagnosis of ONJ will lead to an effective management.
We also increasingly realize that the collaboration between dental and oncology professionals and patients is the key factor in the prevention and management of ONJ related to the bone targeted medications. Research and daily clinical practice have shown, beyond any doubt, that "education" can improve this multidisciplinary, successful collaboration.
The ecancer online learning ONJ module offers a unique tool for oncology training and education, always at a real time setting. Basic Knowledge on the mechanism of action of the bisphosphonates and denosumab, the updated clinical presentation of ONJ and the most recently described early diagnostic clinical criteria, the prevention, risk factors and effective management of ONJ are presented, through a successful multidisciplinary collaboration. Ecancer educational module on ONJ offers the appropriate knowledge to achieve ONJ prevention, early diagnosis and successful management, allowing for a successful BTAs protection of the bones in osteoporosis and in the cancer setting.
*E*cancer online educational module for ONJ was based on a program which evaluated the potential of online communities in knowledge measuring \[[@ref8]\]. To our knowledge there are no other tools for online ONJ education and this was one of the reasons why the module was created. Online learning can be accessed anywhere at any time. We can keep up with the increasing oncology information within a safe setting. No delays in knowledge dissemination and education!
Conclusion
==========
Bone targeting agents, administered in osteoporosis or cancer with bone metastases, protect the skeleton and reduce skeletal-related adverse events, maintaining quality of life. BTAs have been associated with ONJ, a rare, but dose-limiting bone complication.
The ecancer online educational module offers information to clinicians prescribing these drugs. Their collaboration with dental professionals can contribute to improved and efficient prevention, early diagnosis and successful management of ONJ.
Conflicts of interest
=====================
Ourania Nicolatou-Galitis has received consultant fees from AMGEN.
Authors' contributions
======================
Ourania Nicolatou-Galitis wrote the article and Dr Cesar Migliorati edited it.
We would like to thank ecancer.org for developing the elearning modules and Dr Jean Jaques Body, who was the scientific responsible for this osteonecrosis of the jaw module.
{#figure1}
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Computed tomography angiography (CTA) is a simple and noninvasive procedure for the evaluation of the hepatic vasculature. Although conventional angiography can also provide assessment of the hepatic vasculature, the modality is invasive and more technically difficult to perform than CTA. CTA is replacing conventional angiography in the depiction of the normal vascular anatomy and the diagnosis of vascular disorders \[[@B10]\]. However, CTA two-dimensional (2D) images do not provide complete three-dimensional images of the hepatic vascular anatomy. As three-dimensional (3D) reconstruction provides more comprehensive and accurate anatomic information, 3D CTA is a useful method to improve the limitations of the use of 2D images.
Three-dimensional CTA represents an increasingly important clinical tool that is used to diagnose portal hypertension and hepatic vascular disorders. These disorders include the presence of a single or multiple extrahepatic portosystemic shunt, intrahepatic portosystemic shunt, portal vein thrombosis, intravascular tumor extension, and well-developed vascular tumors such as a hepatocellular carcinoma, liver or pancreatic neoplasia. Three-dimensional CTA is also used to evaluate suspected liver disease, and is used for surgical planning \[[@B1],[@B3],[@B9],[@B11]\].
In veterinary medicine, CT portography has been performed in normal dogs and in veterinary subjects with portosystemic shunts \[[@B3]\] to develop the dual-phase CT angiography technique for normal dogs \[[@B11]\]. However, methods of 3D reconstruction and 3D CTA analysis for the canine hepatic vasculature have not been investigated. The objectives of this study are 1) to develop the CTA technique for imaging of the canine hepatic vasculature and 2) to describe the anatomy of the hepatic vasculature with the use of 3D CTA.
Materials and Methods
=====================
Animals
-------
Eight healthy Beagle dogs, ranging from 1 to 4 years old and weighing 6 to 12 kg were used in the study. A 22 G indwelling catheter was placed in a cephalic vein and was connected to a CT power injector (LF CT9000 ADV; Liebel-Flarsheim, USA) by an extension line (Control pressure line; Hyup Sung Medical, Korea).
Contrast media
--------------
Non-ionic iodine contrast media, iohexol (Omnipaque 300 mg Iodine/ml; Amersham Health, UK) was used for cine scans and for enhanced scans. Contrast media was injected by the use of a CT power injector and all injection rates were 2.0 ml/sec.
Helical CT scaning and parameters
---------------------------------
CT angiography was performed using a single slice helical CT scanner (GE CT/e; GE Healthcare, USA). The CT programs for image analysis were as follows. 1) The use of cine CT; 2) the use of a retrospective reconstruction program; 3) the use of a reformatted program (for axial, sagittal, transverse and oblique plane images); 4) the use of a cross section histogram for measurement of Hounsfield units (HU) in the 2D plane; 5) the use of a 3D display of shaded surface display (SSD) and maximum intensity projection (MIP).
Experimental animal preparation for CT scanning
-----------------------------------------------
General anesthesia was performed to avoid motion artifacts and breath holding was induced by hyperventilation. The position of the animals was dorsal recumbency and the heads of the dogs were placed toward the CT gantry.
Experimental design
-------------------
**Unenhanced scan:** Unenhanced scans were performed to determine the location of the cine scan, the scan field range of the enhanced scan and to measure pre-contrast HU values of the aorta (AO), caudal vena cava (CVC), portal vein (PV) and liver. Conditions included a 5 mm thickness, 3 mm interval, 1.5 pitch, 120 kVp and 40\~60 mA in each animal. Scans were started from the cranial aspect of the diaphragm to the caudal aspect of the fourth lumbar vertebra. After the unenhanced scan was performed, the cine scan location, scan field range of the enhanced scan and pre-contrast HU values of each vessel and the liver at the cine scan location were determined.
**Cine scan:** In the cine scan, operating conditions of a 3 mm thickness, 120 kVp, 30\~55 mA, 1.5 sec per rotation scan speed and 50 serial axial images for 78 sec were performed. The cine scan was performed at the site of the well visualized AO, CVC, PV and liver at the thirteenth thoracic vertebra level. This scan was performed in order to obtain time-attenuation curves of the injected iohexol (0.5 ml/kg). With the use of time-attenuation curves, the delay time for helical CT acquisition was obtained. Time-attenuation curves of the AO, CVC and PV were achieved by using a hardwired basic CT program on the CT unit ([Fig. 1](#F1){ref-type="fig"}).
**Enhanced scan:** The enhanced scan was divided into the arterial phase and the venous phase. Arterial phase images were obtained by caudocranial data acquisition and venous phase images were obtained by craniocaudal acquisition to minimize time during the dual phase scan. The enhanced scan was performed by using the following parameters of a 5 mm thickness, 5 mm interval, 120 kVp, 40\~60 mA and 1.3\~1.5 pitch. Time-attenuation data were used to optimize the delay of CTA image acquisition following IV injection of contrast medium to maximize hepatic vessel opacification. A delay time was applied to arterial phase scanning. The venous phase was started directly after the arterial phase scan termination. The contrast medium dose used was 3 ml/kg. All enhanced phase data were reconstructed to 1 mm interval images retrospectively.
Statistical analysis
--------------------
Statistical analysis was performed by the use of SPSS software (SPSS 12.0.0; SPSS, Chicago, IL USA). A one-way ANOVA least significant difference test was applied for quantitative data analysis. The Kruskal-Wallis test and Mann-Whitney U test were applied for qualitative data analysis \[[@B1],[@B2]\].
Results
=======
Pre-contrast HU
---------------
With the use of unenhanced scan images, the HU values of the AO, CVC, PV, and liver were measured in the determined cine scan location ([Table 1](#T1){ref-type="table"}). Regions of interest (ROI) were set at above 80% of the vascular diameter in the center of the vessels but not outside of the vessel outline. In the liver, the ROI was placed in the liver parenchyma while avoiding vessels.
Time-attenuation curves
-----------------------
Time-attenuation data were used to optimize the delay time of the CTA image acquisition following IV injection of contrast medium to maximize hepatic vessel opacification. The time delay required for subsequent scans was defined as the time post-onset of contrast administration to the rise in the vessel HU as compared to the baseline HU plus 20 ([Fig. 2](#F2){ref-type="fig"}). This delay time was applied to the enhanced scans. The delay time of the arterial phase was set at 9 sec ([Table 2](#T2){ref-type="table"}). As the initial times were so close between the CVC and PV that were approximately 18 sec and 22 sec, the delay time of the venous phase was set at 18 sec of the initial time of the CVC.
**Delay time and enhanced scan periods:** In the enhanced scan, the arterial phase scan was started after applying the delay time, and the venous phase scan was started immediately after termination of the arterial phase. The total enhanced scanning time was approximately 67 sec. The consumed scan-time was recorded at each phase ([Table 3](#T3){ref-type="table"}).
**3D reconstruction:** Vascular 3D mapping was performed by use of a hardwired basic CT program on the CT unit. 3D SSD reconstruction was applied to a threshold-based reconstruction technique. The procedure of the threshold-based reconstruction technique is that the 3D threshold is increased in order to select opacified vessels. After applying the 3D threshold, the 3D vascular structure was seen and the background with low HU value was eliminated. With the use of 3D reconstruction, reformatted and MIP images, the anatomical location of each vessel was confirmed ([Figs. 3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"}).
**Quantitative measurement:** With the use of a cross section histogram ([Fig. 5](#F5){ref-type="fig"}), the AO, CVC and PV were measured by phases of the enhanced scans ([Tables 4](#T4){ref-type="table"}-[6](#T6){ref-type="table"}).
Discussion
==========
Imaging methods for the hepatic vasculature include conventional angiography, ultrasonography, and CTA. Conventional angiography for the hepatic vasculature can provide good vascular images. However, the use of the modality is invasive to patients, has a relatively high cost as compared to other methods \[[@B6]\], requires a difficult technique of superselective catheterization for the hepatic arteries \[[@B8]\] and is time consuming. Ultrasound is an expedient method for imaging of the hepatic vasculature, but disadvantages include the disparity of accuracy between sonographers \[[@B3]\] and many factors such as bone, gas and fat that can interfere with the transmission of the ultrasound beam \[[@B2]\].
CTA provides a fast, noninvasive modality for the evaluation of the hepatic vasculature. The use of a helical CT scan with the advanced 3D display technique provides detailed anatomic images of the hepatic vasculature and requires little time. It is also less than one-third the cost of conventional angiography, and is not dependent on the skill of the operator performing the study or on the body habitus of the patient \[[@B6]\].
The most important parameter of the hepatic CTA was the \'time delay\' between the injection of contrast medium and image acquisition. When the delay time is applied to a scan, it permits scanning during maximal enhancement. In this study, the optimal delay time was set at 9 sec in the arterial phase and at 18 sec in the venous phase. This protocol offered good vascular enhancement.
Although the venous delay time was set at 18 sec by the use of a cine scan, the actual real venous phase scan started at 20 sec later for the ideal venous delay time. This was due to the contrast medium injection delay time, the time required for arterial phase scanning and the scan delay in the CT scanner itself between arterial phase scanning and venous phase scanning. In spite of this retardation, it did not affect the image quality.
In all phases of the CT scan, vascular HU values increased as much as the contrast media dose increased. During the arterial phase, there were patterns of increasing vascular HU values, but there was no statistical difference in the HU values despite the dose increase. It was deduced that the opacified difference related to contrast dose did not appear prominent as arteries have characteristics of fast opacifying, deopacifying after contrast media injection and have a relatively smaller size than veins. In the venous phase, there were remarkable opacified differences that were seen related to contrast dosage.
In most CT angiography procedures in humans, an injection rate of various iodine concentrations is used in a range of 1.5\~5 ml/sec \[[@B3]\]. For arterial 3D construction, 5 ml/sec is necessary to achieve a greater intravascular concentration and therefore a higher CT attenuation. Since aberrant hepatic arteries can be relatively small, they need to show sufficient enhancement so that they are not obscured during 3D threshold-based reconstruction \[[@B5]\]. However, in the portal and venous phase, the effect of bolus injection is gradually diminished and a higher injection rate causes a narrow \"temporal window\" (duration of optimal enhancement) \[[@B3]\]. As these factors and with a single channel helical CT limitation, although the arterial bolus effect was decreased, a rate of 2 ml/sec was chosen in this study as the injection rate.
With the use of the MIP technique, vascular anatomy is best depicted when there is a large difference between the attenuation values of vessels opacified by use of contrast agent and the surrounding tissues. However, MIP lacks depth orientation, and the technique is not as capable to display complex anatomy, especially when overlapping vessels are present \[[@B10]\].
Traditional helical single-slice CT scanners are still limited in the ability to image large volumes during a single breathhold and to provide adequate spatial resolution crucial for CT angiography \[[@B6]\]. In this study, due to the limitation of the use of a single channel helical CT scanner, a wide slice thickness and narrow scan range including the liver and the full vascular structures was used. This limitation has prompted the development of faster multidetector helical CT scanners (MDCT) that can cover an extensive volume quickly with excellent spatial resolution \[[@B6]\]. The use of MDCT can overcome the limitations of hepatic CTA that occur with the use of a single channel CT scanner.
In conclusion, 3D CTA has been shown as a useful method for the evaluation of the canine hepatic vasculature.
{#F1}
{#F2}
{#F3}
{#F4}
{#F5}
######
Pre-contrast Hounsfield unit values of the aorta (AO), caudal vena cava (CVC), portal vein (PV) and liver
######
Initial and peak intensifying time and Hounsfield unit (HU) values
n = 29, Contrast media dose = 0.5 ml/kg, psi = 47.44 ± 14.34. All data represent mean ± SD.
######
Delay time and consumed scan time of the enhanced scan
n = 29. All data represent mean ± SD.
######
Hounsfield unit values of the arteries as measured in the arterial phase
^a,b^There is statistical significance between a and b within columns (*p* \< 0.05). All data represent mean ± SD. Group A = 2 ml/kg; Group B = 3 ml/kg; Group C = 4 ml/kg. AO = aorta; CA = celiac artery; HA = hepatic artery; GD = gastroduodenal artery; RG = right gastric artery; LG = left gastric artery; RB = right hepatic artery branch; LB = left hepatic artery branch.
######
Hounsfield unit values of the portal veins in the venous phase
^\*^There is statistical significance among groups (*p* \< 0.01). All data represent mean ± SD. MP = main portal vein; SV = splenic vein; CPV = caudal portal vein; RLPV = right lateral portal vein; RMPV = right medial portal vein; QPV = quadrate portal vein; LMPV = left medial portal vein; LLPV = left lateral portal vein.
######
Hounsfield unit values of the hepatic veins in the venous phase
^\*^There is statistical significance among groups (*p* \< 0.05). All data represent mean ± SD. CVC = caudal vena cava; CHV = caudal hepatic vein; RLHV = right lateral hepatic vein; RMHV = right medial hepatic vein; QHV = quadrate hepatic vein; LMHV = left medial hepatic vein; LLHV = left lateral hepatic vein.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-antibiotics-08-00002}
===============
Globally, antibiotic use is increasing, up 65% between 2000 and 2015 \[[@B1-antibiotics-08-00002]\]. The majority of this increase has been caused by rapid expansion of use in low- and middle-income countries (LMICs). Consumption rates in many LMICs are catching up with higher-income countries \[[@B1-antibiotics-08-00002]\] but usage is often unnecessary due to lack of supportive regulation and widespread informal use \[[@B2-antibiotics-08-00002]\]. While access to antibiotics is of clear public health benefit, the inappropriate use of antibiotics is a major driver of antibiotic resistance \[[@B3-antibiotics-08-00002]\] and requires urgent attention. To inform context-appropriate policy in LMICs, we conducted a review of supply side interventions and their impact on unnecessary and inappropriate use of antibiotics in human and animal health systems. The aim was to determine the evidence base for changing the practices of antibiotic prescribers and gatekeepers in LMICs.
Previous systematic reviews have found evidence to support a range of effective and safe interventions to reduce antibiotic prescribing in human health settings. Interventions that combine restrictive and enabling policies have shown better results \[[@B4-antibiotics-08-00002]\]. Interventions that are interactive and multi-faceted have stronger effects, for example education with feedback and monitoring mechanisms \[[@B5-antibiotics-08-00002],[@B6-antibiotics-08-00002]\]. However, the majority of studies included in these reviews are from high-income countries, and within that a limited range of settings (e.g., in-patient prescribing) and methods. One review \[[@B6-antibiotics-08-00002]\] looked specifically for evidence on doctor's prescribing in primary care in upper middle-income countries. However, their review only included eight studies and the authors noted that the quality of evidence was limited. We are not aware of any reviews of prescribing interventions in animal health settings. Taken together, these reviews have found encouraging results but there are questions about how generalizable they are beyond the specific targeted groups (e.g., doctors) and homogenous prescribing contexts (e.g., formal providers in relatively well organised human health systems in high-income countries). LMICs, as defined by World Bank, include a variety of health system arrangements, which makes generalizations difficult. However, in many, the human and animal healthcare contexts can be described as "pluralistic", where people consult a variety of public, private, traditional, biomedical, formal and informal providers, and where the capacity of governments to regulate these providers is variable, and is increasingly recognised as a major concern \[[@B7-antibiotics-08-00002]\]. Effective action is complicated by high rates of infection due to poor preventative public health combined with inconsistent and inequitable access to essential medicines. As such, research on LMICs interventions and contexts is urgently needed and is the focus of this review.
The monitoring of unintended consequences requires particular attention as the supply and distribution of antibiotics has the potential to include competing objectives and is sensitive to interventions not specifically targeting antibiotic use. For example, the use of rapid diagnostic tests to improve malaria diagnosis has led to an increase in antibiotic use in some settings \[[@B8-antibiotics-08-00002],[@B9-antibiotics-08-00002]\].
The review set out to answer the following questions (for human and animal health systems): -What supply-side interventions have been tried in LMICs to reduce inappropriate antibiotic prescribing and sale?-Where have interventions been implemented e.g., country and kind of health care settings?-Which kinds of interventions have shown promise in reducing inappropriate prescribing?-What information exists on the contexts of interventions and attempts to influence antibiotic use and prescribing?
2. Methods {#sec2-antibiotics-08-00002}
==========
We carried out a scoping review to provide an overview of the evidence base, as opposed to narrowing down to aggregate sub-sets of evidence as in a systematic review \[[@B10-antibiotics-08-00002],[@B11-antibiotics-08-00002]\]. In addition to quantitative evaluations of interventions, we include qualitative studies describing the contexts of interventions to enhance understanding of the drivers of prescribing and barriers to reducing prescribing. We carried the review out in five stages: (1) identifying the research question; (2) identifying relevant studies; (3) study selection; (4) charting the data; and (5) collating, summarising and reporting the results \[[@B12-antibiotics-08-00002]\].
2.1. Search Strategy {#sec2dot1-antibiotics-08-00002}
--------------------
We conducted a systematic search guided by a combination of the PICO (Population, Intervention, Comparison, and Outcome) and PICo (Population/Problem, Interest, Context) approach. Full parameters are in [Appendix A](#app1-antibiotics-08-00002){ref-type="app"}. PICO is usually used for quantitative reviews while PICo is used for qualitative reviews. As we were interested in both human and animal antibiotic prescribing, we looked for studies concerning people who supply antibiotics to sick people or animals, e.g., doctors and other healthcare prescribers, drug sellers, informal doctors, pharmacists, community health workers, veterinarians, farmers, and community animal health workers. We looked for any intervention aiming to influence the prescribing or sale of antibiotics (formal and informal), e.g., communication and education, stewardship programmes, treatment algorithms, delayed treatment, alternative treatments, incentives, pricing, packaging, legislation, and peer or community oversight. Outcomes of interest were measured improvement in antibiotic supply (e.g., reduction in unnecessary antibiotic prescriptions and sales, health worker adherence to guidelines, etc.), as well as reported changes in knowledge and attitudes around antibiotic prescribing, health outcomes (improved, unaffected, or adverse), and unintended consequences. Qualitative analysis of the contexts of interventions were also of interest, e.g., implementation of a new policy and descriptions of attempts to influence prescribing.
To identify research in both the biological and social sciences, we searched MEDLINE, SCOPUS and the Cochrane database of clinical trials. We also searched the 3ie database of impact evaluations and the World Organisation for Animal Health (OIE) database. Search terms are included in [Appendix B](#app2-antibiotics-08-00002){ref-type="app"}. [Table 1](#antibiotics-08-00002-t001){ref-type="table"} contains the search results for each database searched. Time and resource constraints meant we limited the date range to studies published during 2000--2017. A number of reviews already covered publications before this cut off \[[@B4-antibiotics-08-00002],[@B5-antibiotics-08-00002],[@B6-antibiotics-08-00002]\], and most studies were published post-2000, so we anticipated the effect of the limitation to be minimal
2.2. Study Selection {#sec2dot2-antibiotics-08-00002}
--------------------
Two research assistants reviewed the abstracts using the pre-defined inclusion and exclusion criteria below. Borderline studies were discussed within the team. [Table 2](#antibiotics-08-00002-t002){ref-type="table"} shows the number of studies included and excluded at each stage of the review process. From the initial figure of 15,142 search results, the review includes 102 studies. Studies of interventions and intervention contexts were separated out.
Inclusion criteria:Reports on an intervention aiming to influence the prescribing and sale of antibiotics (formal and informal), including the use of antibiotics to treat existing and suspected infections or to prevent infection (e.g., antibiotic prophylaxis)Any healthcare setting in a LMIC countryOutcomes: Externally measured change in antibiotic use/sale/prescribingExternally measured change in knowledge and attitudes around antibiotic useSelf-reported change in antibiotic use/sale/prescribingSelf-reported change in knowledge and attitudes around antibiotic useObserved change in the quality of antibiotics prescribed/sold Or for "context studies" Studies that describe the contextual factors that influenced the effectiveness of interventions
Exclusion criteria: -Studies of other ways of addressing antibiotic resistance, e.g., hygiene and infection and prevention control such as vaccines, etc.-Studies assessing adherence to antibiotics or medicine-Studies of environmental transmission of antibiotics and antibiotic resistance-Studies targeting other antimicrobials, and not targeting antibiotic use-Studies reporting on/intervening in patient demand for antibiotics and self-use-Studies evaluating effectiveness of antibiotic treatment in clinical care, e.g., comparing different treatments or ways of administering antibiotics, and reporting on clinical outcomes.-For "context studies", those studies that only described the context and not the implementation or outcome of an intervention, e.g., a situation analysis
We carried out a detailed data extraction on the remaining 102 studies. For the 32 context studies, we extracted qualitative data summarising their findings, and indicating the setting. For the 70 intervention studies, we extracted descriptive information about the setting, intervention design and implementation, study design, the author's assessment of impact and our own.
We looked at intervention effectiveness in more detail in a subset of studies (*n* = 41). Studies that contained adequate information to assess effectiveness were included in this analysis. Two researchers assessed studies for inclusion. Our assessment was based on the extent to which studies contained complete information about the intervention and study design and thus whether necessary information was available and how much confidence we had in their results (e.g., sample sizes, *p*-values, primary outcomes, descriptions of the intervention, and identification of setting). Studies that comprehensively discussed their methods and outcomes, and reported strategies to minimize estimation bias and limitations were included. Studies were excluded if they were missing key study design or analysis information such as the selection and measurement of outcomes, sample sizes, or p-values. Those studies that interpreted results incorrectly or employed inconsistent methods for data collection without justification or explanation were also excluded.
3. Results {#sec3-antibiotics-08-00002}
==========
We present summary statistics of the intervention studies included in the review. [Table 3](#antibiotics-08-00002-t003){ref-type="table"} shows the number of studies included from each country. The largest number of studies are from China (25 out of 70 studies), followed by India (6 studies). The high number of studies in China may be in response to the major health reforms that took place there, e.g., in 2009, which saw the implementation of Essential Medicine Lists, and, during 2011--2013, when there was a national campaign on improving the use of antibacterial drugs. There are two cross-country studies---one in Brazil and Mexico, and the other in Vietnam and Thailand. The rest of the studies were conducted in a single-country setting.
3.1. Intervention Settings {#sec3dot1-antibiotics-08-00002}
--------------------------
We found no interventions addressing antibiotic prescribing in animal health. All intervention studies were from human healthcare settings. The overwhelming majority of the reviewed studies evaluated interventions at public prescribing facilities rather than private (see [Figure 1](#antibiotics-08-00002-f001){ref-type="fig"}). This trend is particularly pronounced among studies conducted in primary care settings, where 15 out of 22 studies examined the public sector. The exception was for pharmacies where seven out of nine studies were conducted in the private sector. This is likely because pharmacies and drug shops are usually private enterprises. In the hospital settings, 16 and 4 studies were conducted in public and private hospitals, respectively, while in the remaining 16 it is unclear. This was due to the fact that studies mentioned hospitals' names but we were unable to judge whether the hospitals concerned are public or private from available sources. Three other studies are not included in the figures: two were from mixed public and private sector settings, and one was not clear.
There was also a majority of studies from urban areas, as shown in [Figure 2](#antibiotics-08-00002-f002){ref-type="fig"}. In total, 41 out of 70 studies looked at interventions in urban areas. In particular, studies conducted at hospitals and primary care mostly evaluated urban settings. Two of the studies carried out in mixed health care settings were from rural areas, while the remaining one study addressed both rural and urban areas.
3.2. Types of Intervention {#sec3dot2-antibiotics-08-00002}
--------------------------
[Table 4](#antibiotics-08-00002-t004){ref-type="table"} presents the types of intervention conducted in different settings in a "gap map" format. Interventions were classified into six broad categories: norms and standards, knowledge interventions, decision support, supply chain, economic, and organisational/management systems. Blank cells indicate that we found no evidence of an intervention being evaluated in that setting. The majority of interventions have only been studied 1--3 times, and often in only one type of setting. The evidence base is uneven, for example, there are 35 hospital-based studies and only nine in pharmacies. Only educational interventions and the essential medicine policy have been evaluated in all settings.
[Table 4](#antibiotics-08-00002-t004){ref-type="table"} reveals a number of gaps in the evidence base though it should be noted that not all forms of intervention are as relevant for each setting. "Audit/Feedback" and "Stewardship Programme" are most often implemented in hospitals in comparison to other settings. This might reflect the ease of monitoring and management in hospital settings, which are arguably more "closed", as well as the availability of expertise (e.g., infectious disease specialists) to perform key supportive roles. All five studies conducted on essential medicines policy in primary care setting come from China where there was major national policy reform. Although essential medicine policies are used widely, caution is needed when considering the generalisability of those studies that report on a specific policy moment and context.
3.3. Intervention Impact {#sec3dot3-antibiotics-08-00002}
------------------------
We examined impact in a subset of 41 studies (out of 70), selected on the basis of their reporting completeness. Studies were classified as having positive, negative, mixed, or no effect, on the basis of whether they reported improved or deteriorated prescribing rates, a combination, or no change, respectively. Studies reporting decreases in antibiotic prescribing or increased adherence to appropriate guidelines (e.g., choice of antibiotic, timing, dose, and duration of treatment) were classified as positive. Studies reporting increased antibiotic prescribing rates or decreased adherence to guidelines (e.g., increased inappropriate antibiotic use, and use of restricted antibiotics) were classified as negative. [Table 5](#antibiotics-08-00002-t005){ref-type="table"} shows the number of studies reporting each category of impact for each type of intervention.
Twenty-one studies show positive effects of the interventions, thirteen studies document mixed effects, three studies show no effects and four studies indicate negative effects. Studies that had mixed results were those that reported both positive and negative impacts on different prescribing indicators and thus could not be categorised as wholly positive or negative. It should be noted that sometimes this may reflect more complex analysis than the studies reporting positive (or negative) results. The higher number of studies reporting positive or mixed results than negative or no results may be indicative of publication bias. [Table 5](#antibiotics-08-00002-t005){ref-type="table"} includes all intervention types identified by the review so there are some complete rows with no studies, e.g., supply chain interventions where the evidence was found and included in [Table 4](#antibiotics-08-00002-t004){ref-type="table"}, but excluded here because of reporting deficiencies.
Overall the evidence base is limited. Results are scattered and number of studies in each category of result and intervention type is small; four being the maximum number of studies reporting positive or mixed results for any intervention type. Out of the studies with positive results, the strongest evidence is clustered around knowledge interventions, and there are four studies for audit/feedback interventions and four for education/feedback interventions. Audit/feedback studies are those which introduced primarily audit interventions (e.g., reviews of prescribing patterns according to guidelines) combined with feedback mechanisms (e.g., meetings and reports). Education/feedback studies are those which introduced educational interventions (e.g., training sessions or courses and role play) combined with feedback (e.g., meetings and group discussion). There are five positive evaluations of stewardship interventions. Stewardship studies are those which combined policy review and change (e.g., new targets, development of new guidelines, and incentives/disincentives), the creation of antimicrobial/antibiotic resistance committees or working groups, education, audit and monitoring. Out of studies reporting mixed results, four studies evaluated essential medicine policies as well as public reporting (e.g., monitoring prescribing patterns and posting these publicly in/around health settings). Studies documenting no or negative effects evaluated accreditation, diagnostics, essential medicines policy, education, health insurance, and pricing intervention. More details of intervention types are in [Appendix C](#app3-antibiotics-08-00002){ref-type="app"}.
Given that past reviews have found evidence suggesting that multifaceted interventions have greater impact than single interventions, we checked if similar trends were observable in our studies. We classified 21 studies as using multiple/mixed intervention strategies (e.g., education with feedback or monitoring, or stewardship programmes which are multi-strategy) and 19 as using single change pathways (e.g., one intervention modality, such as training). In one study, the approach was unclear. This study was a survey of doctor's prescribing practice which included questions on whether they had received training on antibiotic use, however no details of the training were given. [Table 6](#antibiotics-08-00002-t006){ref-type="table"} shows that although both single and multifaceted interventions produced positive and mixed results, the majority of studies reporting mixed or positive results were multifaceted studies (positive and mixed combined = 20 versus 13) and studies evaluating single pathway interventions reported no or negative effects more frequently (negative and no effect combined = 1 versus 6). Indeed, none of the interventions classified as having mixed pathways reported negative results, and only one concluded there was no effect (see [Appendix C](#app3-antibiotics-08-00002){ref-type="app"} for more detail).
3.4. Prescribing and Intervention Contexts {#sec3dot4-antibiotics-08-00002}
------------------------------------------
The review identified 32 context studies, from 18 different countries. One study was multi-country \[[@B13-antibiotics-08-00002]\] while the rest described specific country contexts, which are listed in [Appendix D](#app4-antibiotics-08-00002){ref-type="app"}. These studies described the context of interventions aiming to address antibiotic supply in hospitals, primary care and pharmacy settings, including public, private and informal providers. At least five studies looked specifically at drug shops, community pharmacists and informal village doctors. Only one study describing use of antibiotics in animal health was identified.
### 3.4.1. Knowledge of Antimicrobial Resistance and Appropriate Antibiotic Use {#sec3dot4dot1-antibiotics-08-00002}
A number of studies identified limited knowledge of antimicrobial resistance (AMR) and of appropriate drug use in a range of levels and sectors, for example surgeons in Jordon \[[@B14-antibiotics-08-00002]\], tertiary and primary care doctors in China \[[@B15-antibiotics-08-00002]\], doctors in Laos \[[@B16-antibiotics-08-00002]\], public and private doctors and pharmacists in India \[[@B17-antibiotics-08-00002]\] and drug sellers and pharmacists in Vietnam \[[@B18-antibiotics-08-00002]\]. Bai et al. \[[@B15-antibiotics-08-00002]\] found that, although age and gender did not appear to be associated with knowledge scores in Chinese settings, it was related to level of training, with more senior doctors performing well, and primary care doctors having lower levels of knowledge than tertiary care doctors. In India, urban doctors seemed to have better knowledge of AMR than rural ones \[[@B17-antibiotics-08-00002]\]. Similarly, accreditation was not associated with improved knowledge and prescribing among drug sellers and pharmacists in Vietnam; however, level of training was, which in turn mapped onto urban--rural differences, with rural pharmacists being less qualified than urban ones \[[@B18-antibiotics-08-00002]\]. In contrast, Reynolds and Mckee's \[[@B19-antibiotics-08-00002]\] in depth study of knowledge and practice among Chinese health workers found that awareness of resistance was high (although understanding of mechanisms was limited), but despite this health workers still prescribed inappropriately due to incentives listed in the next Section.
### 3.4.2. Practical Concerns and Diverse Influences on Antibiotic Prescribing {#sec3dot4dot2-antibiotics-08-00002}
Research pointed to structural issues impeding healthcare workers' ability to adhere to standards and guidelines. For example, in Jordan, surgeons reported that appropriate drugs were absent \[[@B14-antibiotics-08-00002]\], as did pharmacists in India \[[@B20-antibiotics-08-00002]\].
Beyond medicine availability, there appear to be myriad more subtle influencing factors. Studies of doctors \[[@B17-antibiotics-08-00002],[@B19-antibiotics-08-00002],[@B20-antibiotics-08-00002]\] identified the following factors as influencing prescribing: (i) inadequate diagnostic facilities; (ii) lack of antibiotic guidelines; (iii) difficulty in observing patient progress (e.g., patients do not return for test results and follow up, especially in poorer areas and in public clinics where they cannot afford tests); (iv) poor intensive care facilities in rural areas; (v) patient demand for quick relief; (vi) perceived patient expectation from past prescriptions; (vii) using up stock; and (viii) fear of losing patients to competition (for private practices). Kotwani et al. \[[@B20-antibiotics-08-00002]\] noted that doctors in the public sector complained they did not have enough time during each consultation to cover patient history in depth or to dissuade patients from wanting antibiotics.
Pharmacists and drug sellers \[[@B17-antibiotics-08-00002],[@B20-antibiotics-08-00002],[@B21-antibiotics-08-00002]\] had a similar list of influences: (i) patient demand; (ii) a belief that cure is through antibiotics; (iii) competition from other pharmacy shops; (iv) antibiotic sales promoting business; (v) the need to use up stock; and (vi) distance and costs of reaching care facilities. Notably, public sector pharmacists in India reported overprescribing antibiotics to deal with stock which was nearing its expiry date \[[@B20-antibiotics-08-00002]\]. They did not return stock as it was a lengthy process and more senior officials pressured them to use up stock. Drug shop owners in Tanzania attributed their unnecessary sales of antibiotics to: customer demand, habit ("mazoea"), following inappropriate health facility prescriptions, the need to make a profit (e.g., sell more expensive medicines), and also the need to use up stock \[[@B22-antibiotics-08-00002]\]. Significantly, for these private sector providers, they were disinclined to refer patients to doctors, as it could result in their patients/customer base losing confidence in them; thus, not selling drugs had short- and potentially long-term negative effects on profit.
An enlightening study from Vietnam \[[@B18-antibiotics-08-00002]\] found that antibiotics were not the most profitable drug for sellers, although it was the most commonly sold drug and the mark up did vary considerably; antibiotics contributed 24% and 18% of total revenue in urban and rural pharmacies, respectively, while herbal remedies contributed 24% and 27%.
### 3.4.3. Industry Influence {#sec3dot4dot3-antibiotics-08-00002}
The influence of the pharmaceutical industry in providing information and incentives looms large \[[@B17-antibiotics-08-00002],[@B23-antibiotics-08-00002],[@B24-antibiotics-08-00002],[@B25-antibiotics-08-00002]\]. In Nigeria, both public and private hospitals' doctors indicated that industry was their primary source of information about drugs \[[@B23-antibiotics-08-00002]\]. Indian doctors and pharmacists reported coming under pressure and being offered incentives by pharmaceutical companies to use or sell drugs, especially newer brands \[[@B17-antibiotics-08-00002]\]. A study of doctors in Pakistan found that private sector doctors were (reportedly) more susceptible to patient demand and to fear of losing patients to other providers than their public sector counterparts, while public sector doctors were more likely to face problems with the availability of drugs \[[@B26-antibiotics-08-00002]\]. In Laos, doctors in provincial settings were more likely to receive information from drug companies than doctors in district and central hospitals \[[@B16-antibiotics-08-00002]\].
In their study of healthcare worker incentives in China, Reynolds and Mckee \[[@B19-antibiotics-08-00002]\] found that some doctors had arrangements with pharmaceutical companies to split the profits of extra sales. This was in addition to the sanctioned cost recovery incentives (pre-2009 Chinese policy reforms).
### 3.4.4. Care and Responsibility {#sec3dot4dot4-antibiotics-08-00002}
Studies identified additional cultures of care that influenced prescribing, suggesting the idea that prescribers and sellers are primarily influenced by stock and profits may be an over simplification. Indian pharmacists said the antibiotics they sold to poor patients was akin to social work \[[@B20-antibiotics-08-00002]\]; and Indian doctors said they prescribed antibiotics as prevention as their patients living with poor sanitation and in unhygienic conditions were susceptible to infection \[[@B27-antibiotics-08-00002]\]. There was also some reluctance for pharmacists to challenge doctor's authority by contesting prescriptions even if they knew they were unnecessary \[[@B20-antibiotics-08-00002]\].
### 3.4.5. Perceptions of Public Perception {#sec3dot4dot5-antibiotics-08-00002}
Many of the studies include health worker's perceptions of the public and their demands. A common perception among doctors is that patients do not know about diseases and desire antibiotics as a result \[[@B17-antibiotics-08-00002]\]. In Nigeria \[[@B23-antibiotics-08-00002]\], doctors say that when the public pay for health services, i.e., in private hospitals, they expect medicines to be in their bills. Providers in both the public and private sector stated that patient demand was a key factor in their prescribing \[[@B15-antibiotics-08-00002],[@B17-antibiotics-08-00002],[@B19-antibiotics-08-00002],[@B20-antibiotics-08-00002],[@B22-antibiotics-08-00002],[@B26-antibiotics-08-00002],[@B27-antibiotics-08-00002],[@B28-antibiotics-08-00002]\]. However, it is not clear what evidence these views are based on and whether they involve misperceptions on the part of providers. In Sri Lanka, Tillekeratne et al. \[[@B28-antibiotics-08-00002]\] found that although patients admitted expecting drugs they were not specifically demanding antibiotics.
### 3.4.6. Perspectives on Interventions {#sec3dot4dot6-antibiotics-08-00002}
Some studies asked participants for suggestions of interventions to improve antibiotic use, which included: creating public awareness, better healthcare provider communication, improved diagnostic support, stricter implementation of guidelines, continuing education, and strengthening of regulations \[[@B17-antibiotics-08-00002]\]. There was an appetite for training \[[@B16-antibiotics-08-00002],[@B29-antibiotics-08-00002]\]. For example, in Laos, a large majority of doctors surveyed were in favour of educational programmes, whereas only 45% of doctors thought guidelines were effective \[[@B16-antibiotics-08-00002]\]; indeed, some doctors thought that guidelines were more of an obstacle than a help (22%). South African medical students expressed a preference for e-learning platforms, but significantly did not appreciate more interactive learning techniques such as "problem-based learning" or "registrar interaction" (i.e., interaction with senior colleagues, in this case trainee specialists) \[[@B30-antibiotics-08-00002]\]. This could reflect poor implementation of these learning modalities but it suggests that clarification is needed of their role in education.
### 3.4.7. Policy Coordination {#sec3dot4dot7-antibiotics-08-00002}
In Bangladesh, the progressive and once highly acclaimed National Drug Policy of 1982 has not been maintained \[[@B31-antibiotics-08-00002]\]; the regulatory institutions do not have power to curb prices or excess use-generics less available, high mark up and sale of antibiotics for cough and fever. Bangladesh's experience serves as an important reminder, for example in relation to China's apparently effective essential drug reform, that there must be sustained investment into the comprehensive implementation of policy especially those which address antibiotic use as part of other agenda's such as universal health coverage. A policy analysis in Namibia \[[@B32-antibiotics-08-00002]\] highlighted inconsistency in national policy frameworks. While treatment guidelines (e.g., HIV, malaria, tuberculosis, etc.) included recommendations about antibiotics there was no overarching guideline for antibiotics; meanwhile, many of the antibiotics included on the essential medicine list were not included in the treatment guidelines, and vice versa. The study noted that lack of coordination between the committees developing these lists had led to confusion with implications for clinical practice. Similarly, in China, doctors complained about the number of new guidelines and no single authoritative source \[[@B19-antibiotics-08-00002]\]. It is relevant that the only international study, a survey of hospital-based stewardship initiatives \[[@B13-antibiotics-08-00002]\], found that major barriers included funding and human resources, suggesting sustainable implementation is a challenge.
### 3.4.8. Unintended Consequences {#sec3dot4dot8-antibiotics-08-00002}
The management of fevers and common illnesses in outpatient settings is a significant problem. A study examining the appropriateness of treatment in Papua New Guinea \[[@B33-antibiotics-08-00002]\] found there was poor compliance with the Integrated Management for Childhood Illness guidelines (ICMI) and that 40% of children were not treated accurately: 29% received antibiotics when they should not have and 11% did not receive antibiotics when they should have done. They observed that the prescription of antibiotics was strongly influenced by the result of rapid diagnostic tests (RDTs) for malaria in feverish patients: in children presenting with criteria for mild pneumonia (who should receive antibiotics), only 40% were treated when the malaria RDT was positive, compared to 76% when the test was negative. This suggests that health workers are primarily following RDT results and ignoring other clinical symptoms, meaning the management of co-morbid bacterial and malarial infections are poorly managed. Several studies point to the inadequacies of apparently simple methods (e.g., fever charts or RDTs) of managing fevers and antibiotic use successfully at the community level \[[@B34-antibiotics-08-00002],[@B35-antibiotics-08-00002]\]. Comprehensive follow up of patients who may have co-morbid infections or who may have been misdiagnosed should be a priority.
### 3.4.9. Animal Health {#sec3dot4dot9-antibiotics-08-00002}
Finally, only one context study covered use of antibiotics in agriculture. Aquaculture farmers in Vietnam were asked about AMR \[[@B36-antibiotics-08-00002]\]. Approximately half of farmers surveyed used antibiotics, primarily as prophylaxis, which they bought directly from manufacturers not through veterinarians. Farmers were not aware of regulations about antibiotic use and their main sources of information was from drug manufactures, including in seminars jointly arranged by the companies and local government.
4. Discussion {#sec4-antibiotics-08-00002}
=============
The review confirms previous studies \[[@B4-antibiotics-08-00002],[@B5-antibiotics-08-00002],[@B6-antibiotics-08-00002]\] that have found that multiplex interventions combining different strategies to influence behaviour tend to have a higher success rate than interventions based on single strategies. Another similarity to other reviews \[[@B4-antibiotics-08-00002]\] is that many of the interventions which worked well combined restrictive and enabling strategies, i.e., educational techniques combined with forms of monitoring. To the best of the authors' knowledge, this is the largest review to document these patterns in LMIC countries.
The inclusion of evidence on prescribing contexts provides important additional insight. These studies highlighted a web of interacting influences on prescriber behaviour that cannot be reduced to simple motivations, such as profit. Instead, health system quality and availability, education, perceptions of patient demand, bureaucratic processes, competition, and cultures of care all play a role.
Physician perceptions of patient demand \[[@B15-antibiotics-08-00002],[@B17-antibiotics-08-00002],[@B19-antibiotics-08-00002],[@B20-antibiotics-08-00002],[@B22-antibiotics-08-00002],[@B26-antibiotics-08-00002],[@B27-antibiotics-08-00002],[@B28-antibiotics-08-00002]\] requires more attention as observations do not always support this \[[@B28-antibiotics-08-00002]\]. Indeed, elsewhere studies that have observed clinical interactions have shown how doctors prescribe irrespective of patient demand \[[@B37-antibiotics-08-00002]\] and that patient satisfaction is not necessarily linked to whether a prescription was received \[[@B38-antibiotics-08-00002]\], although these studies are old. A more recent study from multiple European settings \[[@B39-antibiotics-08-00002]\] found that, while patients frequently hope for and expect antibiotics, it is less common for them to ask for them explicitly. Patients were prescribed antibiotics considerably more often than they asked for them (54% compared to 10%, respectively), and more than was clinically necessary. The same study found that, in general, patients were satisfied with their consultations, whether prescribed antibiotics or not. Assessing the level and influence of patient demand is difficult as research is often based on reported behaviour and perceptions rather than observations.
It is striking the extent to which different providers blame others within the system: health workers blame patients, pharmacists blame doctors and their incorrect prescriptions, and drug sellers blame competitors saying patients will just "go somewhere else". Such blame shifting highlights a more general importance of interactions between people within the health system: pharmacists fear challenging doctors' prescriptions, drug sellers fear losing their customers, and many doctors and pharmacists get a large proportion of their information about antibiotics from drug detailers who also offer incentives. In discussions of interventions to address AMR, the temptation is to focus on awareness raising and changing the behaviour of individuals, be they healthcare workers including prescribers and pharmacists, or patients. However, the stewardship, supply and use of antibiotics in human and animal healthcare is best viewed with a broad systemic lens \[[@B2-antibiotics-08-00002],[@B40-antibiotics-08-00002]\]. This encompasses an appreciation of how the behaviour of individuals is mediated by a range of structural and contextual factors operating at different scales, which can affect individual agency and decision-making.
Scholars have advocated for a consideration of behavioural change interventions within the context of everyday social practices \[[@B41-antibiotics-08-00002]\]; for a whole system approach (at macro, meso and micro levels) to analysing policy outcomes in the health sector \[[@B42-antibiotics-08-00002]\]; for a consideration of formal and informal health markets for drugs and services for both people and animals \[[@B7-antibiotics-08-00002],[@B43-antibiotics-08-00002]\]; for an appreciation of incentive structures in drug supply chains and the effects on over-the-counter drug retailers and informal providers \[[@B44-antibiotics-08-00002]\]; and for an approach to healthcare worker behavioural change interventions that appreciates individual rationales as well as the contextual settings in which specific behaviours are practiced \[[@B45-antibiotics-08-00002]\]. The overall view is also one of decentred governance and regulation, which departs significantly from traditional "command and control" models \[[@B46-antibiotics-08-00002]\]. It is notable therefore that while a majority of studies we identified in this review implemented multiplex interventions, most of these took place within one health setting, e.g., a stewardship programme in a hospital. Very few studies targeted different kinds of healthcare provider and different kinds of health settings. One exception was by Hoa et al \[[@B47-antibiotics-08-00002]\] who attempted an educational programme for all providers in one district, public, private, formal and informal and primary or secondary care. In this example, however, there was a particularly high drop-out rate among informal providers illustrating the challenges of engaging hard to reach groups in interventions. Moreover, important interactions between types of providers are not being addressed and it is rare to find examples of whole system approaches which have been recommended \[[@B2-antibiotics-08-00002],[@B40-antibiotics-08-00002]\]. China provides one of the few examples of system-wide interventions at the national and regional level. There is a need for coordinated action and research across multiple settings and actors throughout and between the human and animal supply chain.
Significantly, our study was unable to identify a single study on interventions that addressed prescribing behaviour for veterinary use of antibiotics, and only one context study on agricultural use. Evidence is emerging that antibiotics critical for human health are being used in animal farming \[[@B48-antibiotics-08-00002]\] including colistin \[[@B49-antibiotics-08-00002],[@B50-antibiotics-08-00002]\]. Research needs to address agricultural use of antibiotics and its overlap with human health system in terms of the antibiotics used, healthcare personnel, and transmission pathways of resistant pathogens. Measures to improve essential medicine use provides a precedent for success \[[@B51-antibiotics-08-00002]\]. The National Action Plan process provides an opportunity to facilitate this. In particular, employing the "One Health" lens \[[@B52-antibiotics-08-00002],[@B53-antibiotics-08-00002]\], taking the intersection among human, veterinary, and environmental health into account, can help address the multi-sector nature of AMR and develop systematic strategies to tackle this challenge.
In addition to the major lack of evidence documenting the effectiveness of interventions to improve veterinary prescribing, this review has identified some gaps in the human evidence. Studies in hospitals were most common, possibly as hospitals are more observable and controllable environments. The overwhelming majority of studies evaluated interventions at public facilities. There was also a majority of studies from urban areas, with less from rural areas, possibly because hospitals are over represented in the data and tend to be in urban areas. Private providers, and in particular those in primary care settings, is another gap. Informal private providers are especially underrepresented. Evidence from LMICs suggests that rural and/or poor people extensively rely on healthcare services provided by informal practitioners and suppliers of drugs \[[@B54-antibiotics-08-00002],[@B55-antibiotics-08-00002]\]. Therefore, understanding the antibiotic prescribing practices of the informal providers, and improving them, has important implications for health of many marginalized people. There were no interventions in this review involving drug retailers or the pharmaceutical companies despite their prominent role in the contextual studies.
Public understanding and demand is a critical piece of the puzzle and requires action but it must be based on accurate understandings of lay knowledge and demand otherwise messaging and educational interventions will fail. Research is needed into public views and their influence on demand and prescriber behaviour. This should form part of a holistic view of supply and demand.
Our study has a number of limitations. While the literature search was extensive, it may not be exhaustive and studies prior to 2000 were not included. The large number of heterogeneous studies meant we were unable to assess quality of the evidence in detail and we did not compare or combine the effects of selected studies. Further research is needed to assess the strength and magnitude of effects for promising interventions. A majority of studies (*n* = 20) reported positive effects while far fewer reported mixed, negative or no effects (*n* = 13, 4 and 3 respectively). This indicates that publication bias may be an issue. Nevertheless, the review was able to identify trends and gaps in the evidence base.
5. Conclusions {#sec5-antibiotics-08-00002}
==============
The evidence collected in this review comes from a range of health care settings, for example hospitals, primary care, pharmacists or drug shops, and from interventions targeting different types of health providers including doctors, nurses, and drug sellers. This review has identified evidence on interventions to improve antibiotic use among providers in LMIC settings, which, in the authors' opinion, has not been well represented in previous reviews. The review found that multiplex interventions that combine different strategies to influence behaviour tend to have a higher success rate than interventions based on single strategies. However, the evidence base is uneven with hospital and urban contexts over-represented for interventions. There is much less evidence on private providers, especially in primary care settings. Informal private providers who play a major role in drug distribution in LMICs are especially underrepresented. Furthermore, there were no interventions involving drug detailers or the pharmaceutical companies despite their prominent role in the contextual studies. Strikingly, no study was identified that addressed veterinary prescribing of antibiotics.
Evidence on prescribing contexts highlights a web of interacting influences on prescriber behaviour including health system quality and availability, education, perceptions of patient demand, bureaucratic processes, profit, competition, and cultures of care. These contextual studies underscore the importance of interactions between different people within the health system. Although a majority of studies implemented multiplex interventions, most of these took place within one health setting, e.g., a stewardship programme in a hospital. Very few studies targeted different kinds of health provider and interactions across different kinds of health setting. There is an urgent need for coordinated multi-actor studies including multiple settings, by taking a One Health approach, including agricultural settings, and actors throughout the supply chain.
the authors would like to thank Aditi Bhonagiri and Tanvir Ahmed for their research assistance conducting searches and reviewing manuscripts.
Conceptualization, A.W., A.E. and H.M.; Methodology, A.W. and A.E.; Analysis, A.W., A.E. and H.M.; Writing---Original Draft Preparation, A.W.; and Writing---Review and Editing, A.E. and H.M.
This research was commissioned by the Wellcome Trust.
The authors declare no conflict of interest.
antibiotics-08-00002-t0A1_Table A1
######
Search parameters.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Parameter Definition
-------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Population People who supply antibiotics to sick people or animals, e.g., doctors and other healthcare prescribers, drug sellers, informal doctors, pharmacists, community health workers, veterinarians, farmers, community animal health workers
Problem Unnecessary or inappropriate use of antibiotics
Intervention Any intervention aiming to influence the prescribing or sale of antibiotics (formal and informal), e.g., communication and education, stewardship programmes, treatment algorithms, delayed treatment, alternative treatments, incentives, pricing, packaging, legislation, peer or community oversight.
Comparison Not applicable. The review is not limited to controlled or comparative research designs. Studies which report on interventions with or without comparison groups are included
Context Hospital, outpatients, community, online, formal and informal systems in LMICs only.
Outcome Primary: Measured improvement in antibiotic use (e.g., reduction in unnecessary antibiotic prescribing, improved adherence to guidelines)\
Other outcomes could include: Reported changes in knowledge and attitudes around antibiotic use, e.g. health outcomes (improved, unaffected, or adverse), levels of antibiotic resistance and unintended consequences
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
**Search Terms**
Bibliographic database searches combined the following sets of terms: (1)Activity/Outcome of interest: -antibiotic AND (prescribing OR sale OR prescribing OR prescription OR selling OR availability OR consumption or dosage OR use)(2)Type of intervention: -"Behaviour change"-stewardship-drug policy-drug financing-drug committee-Marketing / social marketing-Communication-Outreach-"community engagement"-Evaluation-Training-Guideline\$-Algorithm-curriculum-"mass media"-Information-Television-Radio-Newspapers-Booklets-Pamphlet-Posters-Leaflets-Advocacy-Advice-Fines-Incentive\$-"financial incentive\$"-"financial intervention\$"-Fees-formularies-Tax-Taxation-Prohibit-Prohibition-Ban-Regulation-Legislation-Reform-Restrict-Restriction-Reminder\$-"Peer supervision"-Subsidy-Subsidi\$ation-Peer support-Reward\$-Targets\$-Feedback-Planning-Oversight-"Goal setting"-Audit\$-Monitoring-Intervention OR Impact OR Pilot OR evaluation
Descriptions of interventions (*n* = 41) and factors associated with their effectiveness.
**Norms and Standards**
[Guidelines]{.ul} were found to be effective in hospital settings, either when they were locally developed or formed part of a comprehensive national reform programme. In the study by Zhang et al. \[[@B56-antibiotics-08-00002]\], five Chinese hospitals reduced their antibiotic consumption after the introduction of national guidelines. However, very little information is provided about their implementation in hospitals and what strategies may have supported it. Although the specificity of Chinese policy making contexts (e.g., a backdrop of major health and economic reforms, and a relatively strong government) must be taken into account when considering generalisability, the study does provide observational evidence (measured two years after their introduction) that guidelines can be part of effective response to the overuse of antibiotics. Ristic et al. \[[@B57-antibiotics-08-00002]\] recorded a positive effect of new guideline for prophylactic use of antibiotics in Caesarean deliveries in a Serbian hospital. These guidelines were locally developed and educational sessions accompanied their implementation suggesting a degree of ownership.
[Public reporting]{.ul} interventions produced mixed results. Public opinion/expectations and peer comparison appear to influence prescribing behaviours but in contrasting ways depending on the disease, patient cohort and doctor's ranking. Liu et al. \[[@B58-antibiotics-08-00002]\], Tang et al. \[[@B59-antibiotics-08-00002]\], Tang et al. \[[@B60-antibiotics-08-00002]\], and Yang et al. \[[@B61-antibiotics-08-00002]\] reported effects of public reporting in primary care settings in China. Data on antibiotic prescribing (percentage of prescriptions including antibiotics or injections, and cost) were collected and individual doctors and facilities were compared and rankings were published publicly, e.g., on posters outside facilities. The studies note differences in the effect of public reporting between diseases and age groups according to public perception about appropriateness of prescribing for these categories of illness or people. The intervention also influenced doctors differently according to their ranking, and, in the study by Tang et al. \[[@B60-antibiotics-08-00002]\], doctors categorised as "low prescribers" did not change their habits, even though they could have reduced usage further. These findings indicate that social norms around prescribing are interdependent and must be considered otherwise they can undermine even apparently simple interventions.
[Restricted over-the-counter sales:]{.ul} Santa-Ana-Tellez et al. \[[@B62-antibiotics-08-00002]\] and Moura et al. \[[@B63-antibiotics-08-00002]\] evaluated the effect of controlling over-the-counter sales of antibiotics at pharmacies. National policies imposed financial penalties and closure of businesses if antibiotics are prescribed without prescription. These studies find mixed results but the implementation and reasons for the mixed results (e.g., reductions in Mexico but not Brazil, or in urban or private settings rather than rural or public sector) are not examined in detail.
[Accreditation]{.ul} on its own is an insufficient motivator and regulator of antibiotic sales in private sector drug shops. Minzi et al. \[[@B64-antibiotics-08-00002]\] compared dispensing practices at certified drug sales outlets to non-certified outlets. To gain accreditation, drug shops took part in a training to meet specified quality criteria. After accreditation, the shop would then be allowed to sell a limited range of prescription-only essential medicines and to receive additional support through regular supervision, refresher training, and commercial incentives such as training in business skills and access to microfinance. However, adherence to prescribing guidelines was no better in accredited shops than non-accredited shops; indeed, accredited shops were more likely to prescribe antibiotics without prescription and less likely to refer patients appropriately. The authors noted that many of the staff who had received training and gone through the accreditation process were no longer working in the shops, limiting the retention of institutional knowledge from training interventions.
**Knowledge Based Strategies**
A number of interventions began with training sessions but supplemented them with additional components.
[Audit and feedback:]{.ul} Successful interventions involved repeated feedback combined with supportive face-to-face educational strategies and monitoring, and similar approaches demonstrated effectiveness in both hospitals and primary care settings. Awad et al. \[[@B65-antibiotics-08-00002]\], Eltayeb et al. \[[@B66-antibiotics-08-00002]\], Messina, van den Bergh and Goff \[[@B67-antibiotics-08-00002]\], and Shen et al. \[[@B68-antibiotics-08-00002]\] evaluated the effects of auditing prescribing behaviour and providing feedback in hospitals and primary care settings. Interventions spanned between six months and one year. All interventions aimed at appropriate antibiotic prescribing and included combinations of: audit and feedback, expert review, regular reporting to management/superiors, reminders, recommendations and discussion of inappropriate prescribing, educational seminars and face-to-face training sessions. Awad et al. \[[@B65-antibiotics-08-00002]\] and Eltayeb et al. \[[@B66-antibiotics-08-00002]\] tested different combinations of intervention and found that audit and feedback alone was less effective than when combined with other educational and monitoring components.
[Education and feedback]{.ul}: To different degrees, Apisarnthanarak et al. \[[@B69-antibiotics-08-00002]\], Bantar et al. \[[@B70-antibiotics-08-00002]\], Kafle et al. \[[@B71-antibiotics-08-00002]\] and Opondo et al. \[[@B72-antibiotics-08-00002]\] combined education, support and monitoring into multi-faceted interventions. Components included: Theoretical and practical training sessions, tools (e.g., clinical guidelines and job-aids), management processes (e.g., order/request forms), regular supervision from senior colleagues, opportunities to discuss implementation of guidelines, facilitators/champions, and face-to-face and peer-group discussion and feedback on performance. Three were hospital based and one was in a primary care setting. These interventions had education at their core but combined knowledge with restrictive and enabling approaches.
[Education/feedback/regulation]{.ul}: Chuc et al. \[[@B73-antibiotics-08-00002]\] took a regulatory focused approach to their intervention in pharmacies in Vietnam, but combined it with education and peer-support mechanisms. The intervention involved: visits from inspectors to assess compliance with prescription drug regulation, educational sessions, and the identification of peer leaders to devise peer influence strategies. As above, successful interventions combined restrictive and enabling approaches, and importantly demonstrated effectiveness in private sector pharmacies. Notably, there was considerable emphasis on formal regulatory institutions, including inspections, which would require adequate resources to be effective over sustained periods of time.
[Education]{.ul}: Training can be effective, but there is a gap between actual practice and reported practice, with doctors underestimating or underreporting antibiotic prescribing. Two studies examined straightforward training interventions. Shrestha et al. \[[@B74-antibiotics-08-00002]\] evaluated a five-day training course on new WHO lung health prescribing guidelines for primary care settings. Sun et al. \[[@B75-antibiotics-08-00002]\] assessed knowledge, attitude and practice of doctors in China. Doctors who had attended training reported that they were less likely to prescribe antibiotics. However, a separate analysis of prescriptions revealed high levels of antibiotic prescribing. Esmaily et al. \[[@B76-antibiotics-08-00002]\] assessed a two-day training course for primary care doctors on the principles of prescribing and found no improvement, although the intervention was not extensive.
**Decision Support**
[Algorithms:]{.ul} There is high quality evidence showing that algorithms can improve patient management and reduce antibiotic prescribing in paper or electronic form. However, more specific diagnostic and case management approaches must be balanced with comprehensive follow up of patients, otherwise cases could be missed. Shao et al. \[[@B77-antibiotics-08-00002]\] evaluated a new algorithm to assist with the integrated management of childhood illnesses. The quasi-experimental study documents a large and statistically significant improvement in adherence to guidelines between the control (usual practice) and intervention group. The study is conducted in primary care settings. Rambaud-Althaus et al. \[[@B78-antibiotics-08-00002]\] evaluated the effect of case management algorithms that assist with diagnosis and integrated management of common illness among children in primary care settings. They compared a control group to two different methods of algorithm support: paper-based and electronic (smart phone). While both paper and smartphone-based algorithms are shown to decrease antibiotic use, they do so without being able to identify which patients do need antibiotics and so compared to the control there were large numbers of patients who needed antibiotics but who did not receive them. In effect, the control arm is highly sensitive but not specific, whereas the intervention arms are more specific but less sensitive. Patient follow up must be improved and is relevant to addressing resistance and tackling common health problems.
[Diagnostics:]{.ul} Misdiagnosis of bacterial diseases as flu, and misguided use of antibiotics for flu, can be reduced using a rapid influenza test in outpatient settings. Bhavnani et al. \[[@B79-antibiotics-08-00002]\] evaluated a rapid diagnostic test for influenza in outpatients in rural Thailand and found that a negative influenza result was associated with reduced antibiotic use. However, the scale of the reduction was modest, suggesting that comprehensive reductions will require more than improved diagnostics.
**Economic Strategies**
[Financial incentives]{.ul}: There is evidence that comprehensive changes to financial incentive structures can have positive impacts and be achieved within public primary care settings. Yip et al. \[[@B80-antibiotics-08-00002]\] evaluated the effects of changing from a fee-for-service to a performance-based system in 28 primary care settings in China. Operating within China's public health insurance system, 30% of each facilities' budget was withheld, and they were subject to a performance review. Based on the results of the review, the money that had been withheld was shared among the top performing facilities. The study documents decreased antibiotic use, mostly due to reduced injectable antibiotics. Centralised management was important to the effectiveness of financial incentives in this case.
**Organisational/Management Strategies**
[Stewardship programme]{.ul}: Comprehensive stewardship campaigns using multiple approaches resulted in substantial reductions in antibiotic use and sales in both hospitals and pharmacies. Kallel et al. \[[@B81-antibiotics-08-00002]\], Ma et al. \[[@B82-antibiotics-08-00002]\], Ozkurt et al. \[[@B83-antibiotics-08-00002]\], and Zhang et al. \[[@B84-antibiotics-08-00002]\] evaluated effects of stewardship programme at secondary or tertiary hospitals while Marković-Peković et al. \[[@B85-antibiotics-08-00002]\] studied effects at pharmacies. The interventions involved: The formation of expert interdisciplinary groups to provide oversight and guidance, knowledge dissemination, clinical training, academic engagement (e.g., participation in conferences and seminars), monitoring, feedback, data management and verification, the development of new guidelines and targets and financial penalties. In the study by Ma et al. \[[@B82-antibiotics-08-00002]\], doctors not meeting the new standards had their salary deducted, and, in the study by Marković-Peković et al. \[[@B85-antibiotics-08-00002]\], pharmacists were fined for selling antibiotics over the counter. This latter study with pharmacists also includes public awareness campaigns and policy coordination, as part of a larger national stewardship effort. However, the authors also found that pharmacists gave less oral or written advice alongside antibiotic sales, which they suggested could indicate an attempt to obscure banned practices. Increased restriction on sales could prompt sellers to sell drugs in more underhanded or unsafe ways. Magedanz et al. \[[@B86-antibiotics-08-00002]\] addressed a stewardship programme and in particular an audit-and-feedback component focusing on restricting certain classes of antibiotic. This resulted in uneven patterns of use, with success is reducing use of some antibiotics seemingly driving increased use of others. While restricting classes of antibiotics can be effective especially in the short term, it may drive increased use for other drugs which may not be sustainable over the longer term when reductions in usage across the board are required.
[Essential medicine policies]{.ul} had mixed effects and highlight the need to integrate action on antibiotic use with other health system initiatives. Ding et al. \[[@B87-antibiotics-08-00002]\] and Wei et al. \[[@B88-antibiotics-08-00002]\] addressed the effects of China's essential medicines policy on levels of antibiotic use in hospitals while Song et al. \[[@B89-antibiotics-08-00002]\] and Yao et al. \[[@B90-antibiotics-08-00002]\] studied its effects in primary care settings. Essential medicine policies are multifaceted involving: establishing a list of essential drugs, supply chain measures and price control. Results were mixed across regions, urban--rural settings and different facilities, depending on implementation. Use of some antibiotics increased or decreased depending on their inclusion on the lists, as did cost of treatment. Some interventions coincided with other initiatives (e.g., hospital-level stewardship programme) which made it hard to separate effects, including from wider health system reform (e.g., "no mark-up" policy) and delivery (e.g., health insurance programmes) and could be responsible for changes in use. Yang et al. \[[@B91-antibiotics-08-00002]\] evaluated China's national essential medicine policy and found that, although availability and cost of essential drugs increased, there was no reduction in the number of antibiotics prescribed. This is likely due to the fact that the policy focuses on extending access to quality and reasonably priced drugs and not explicitly on prudent use of drugs (see below). Essential medicine lists provide a policy framework for implementing change at scale but prudent use of antibiotics needs to be considered in their design, otherwise they will not lead to significant reductions in use.
**Studies with Negative Results**
Four studies documented increased rates of antibiotic prescribing or increased levels of inappropriate antibiotic use. The interventions associated with these negative findings were Nigeria's essential drug policy \[[@B92-antibiotics-08-00002]\], China's national health insurance policy \[[@B93-antibiotics-08-00002]\], changes in fee structure in Nepal \[[@B94-antibiotics-08-00002]\] and rapid malaria diagnostics \[[@B95-antibiotics-08-00002]\]. All interventions were conducted at the national level and, with the exception of the malaria RDT, their focus was on improving access to medicines rather than aiming to reduce antibiotic use or to improve antibiotic prescribing practices (hence their authors would not view them as unsuccessful or as having "negative" results). We have described these separately to emphasise the need to align policies and consider unintended consequences. In the case of the Malaria RDTs \[[@B95-antibiotics-08-00002]\], improvements in diagnosing malaria were not matched by improvements in the diagnosis of bacterial disease. An unintended consequence of malaria RDTs is that the reductions in unnecessary use of antimalarial drugs have been offset by increases in antibiotics prescription in cases where the RDT is negative for malaria. Unless action on extending access is coordinated with action to improve use there is potential for interventions to be in conflict.
List of context studies and countries -Bangladesh \[[@B31-antibiotics-08-00002]\]-China \[[@B15-antibiotics-08-00002],[@B19-antibiotics-08-00002],[@B29-antibiotics-08-00002]\]-India \[[@B17-antibiotics-08-00002],[@B20-antibiotics-08-00002],[@B27-antibiotics-08-00002],[@B34-antibiotics-08-00002],[@B96-antibiotics-08-00002],[@B97-antibiotics-08-00002]\]-Jordan \[[@B14-antibiotics-08-00002]\]-Laos \[[@B16-antibiotics-08-00002]\]-Mexico \[[@B98-antibiotics-08-00002]\]-Namibia \[[@B32-antibiotics-08-00002]\]-Nepal \[[@B24-antibiotics-08-00002]\]-Nigeria \[[@B23-antibiotics-08-00002]\]-Pakistan \[[@B25-antibiotics-08-00002],[@B26-antibiotics-08-00002]\]-Papua New Guinea \[[@B33-antibiotics-08-00002]\]-South Africa \[[@B30-antibiotics-08-00002]\]-Sri Lanka \[[@B28-antibiotics-08-00002]\]-Sudan \[[@B21-antibiotics-08-00002]\]-Tanzania \[[@B22-antibiotics-08-00002],[@B35-antibiotics-08-00002],[@B99-antibiotics-08-00002],[@B100-antibiotics-08-00002]\]-Thailand \[[@B101-antibiotics-08-00002]\]-Turkey \[[@B102-antibiotics-08-00002],[@B103-antibiotics-08-00002]\]-Vietnam \[[@B18-antibiotics-08-00002],[@B36-antibiotics-08-00002]\]
{#antibiotics-08-00002-f001}
{#antibiotics-08-00002-f002}
antibiotics-08-00002-t001_Table 1
######
Search results.
Database Result
---------------------------------------------------------- --------
Scopus 20,083
Cochrane Central Register of Controlled Trials (CENTRAL) 3823
3ie Impact Evaluations 26
World Organisation for Animal Health (OIE) 0
Total 32,066
Total after duplicates removed 17,716
Total 2000--2017 15,142
antibiotics-08-00002-t002_Table 2
######
Selection process.
Stages
------------------------------------------------------ --------
Abstract review 15,142
Excluded (did not meet inclusion criteria) 14,734
Included- 408
Full text review 408
Excluded---did not meet inclusion criteria on review 289
Duplicates (e.g., studies reporting same data) 4
Full text not available/not in English 13
Included---context studies 32
Included---intervention studies 70
Total---context and intervention 102
antibiotics-08-00002-t003_Table 3
######
Country settings.
Country, by World Health Organisation Region
---------------------------------------------- ----
African region 15
Kenya 3
Mali 1
Malawi 1
Nigeria 3
South Africa 2
Sudan 2
Tanzania 3
Region of the Americas 5
Argentina 1
Brazil 2
Brazil and Mexico 1
French Guiana 1
European region 7
Republic of Srpska, Bosnia and Herzegovina 1
Serbia 2
Turkey 4
Eastern Mediterranean region 3
Pakistan 1
Iran 2
South-East Asian region 13
India 6
Indonesia 1
Nepal 3
Thailand 2
Thailand and Vietnam 1
Western Pacific Region 27
China 25
Vietnam 2
Total 70
antibiotics-08-00002-t004_Table 4
######
Kinds of intervention, by setting.
Type of Intervention Hospital Primary Care Pharmacies Mixed Settings Total
------------------------------------------- ---------- -------------- ------------ ---------------- -------
Norms and standards (formal and informal)
Accreditation 1 1
Guidelines 3 3
Public reporting 4 4
Restrict over-the-counter sales 2 2
Prescription control 1 1
Knowledge
Audit/Feedback 8 2 10
Education 2 3 2 2 9
Education/Feedback 3 2 5
Education/Feedback/Regulation 2 2
Information 1 1
Decision support
Algorithms 1 2 3
Diagnostics 2 1 3
Supply chain
Decentralisation of Supply 1 1
Drug Delivery 1 1
Economic
Financial incentives 1 1
Pricing strategy 1 1
Health Insurance 2 1 3
Organisational/management systems
Essential Medicine Policy 2 5 1 8
Stewardship Programme 10 1 11
Total 36 22 9 3 70
Number of studies: .
antibiotics-08-00002-t005_Table 5
######
Summary of intervention impacts.
Type of Intervention Reported Impact Total
------------------------------------------- ----------------- ------- --- --- ----
Norms and standards (formal and informal)
Accreditation 1 1
Guidelines 2 2
Public reporting 4 4
Restrict over-the-counter sales 2 2
Prescription control 0
Knowledge
Audit/Feedback 4 4
Education 2 1 4
Education/Feedback 4 1 4
Education/Feedback/Regulation 1 1
Information 0
Decision support
Algorithms 1 1 2
Diagnostics 1 1 2
Supply chain
Decentralisation of Supply 0
Drug Delivery 0
Economic
Financial incentives 1 1
Pricing strategy 1 1
Health Insurance 1 1
Organisational/management systems
Essential Medicine Policy 4 1 1 6
Stewardship Programme 5 1 6
Total 21 13 4 3 41
Number of studies .
antibiotics-08-00002-t006_Table 6
######
Results, by intervention approach.
Reported Results Intervention Pathway Total
------------------ ---------------------- ------- ----
Positive 8 12 20
Mixed 5 8 13
Negative 4 0 4
No effect 2 1 3
Total 19 21 40
| {
"pile_set_name": "PubMed Central"
} |
AGE hypothesis {#s2}
==============
The Maillard or advanced glycation end product (AGE) hypothesis on the pathogenesis of diabetic vascular complications proposes that increased protein glycation during hyperglycemia and accelerated accumulation of AGEs on long-lived tissue proteins are fundamental processes underlying the development of diabetes complications ([@B1],[@B2]). The terms "autoxidative glycosylation" and "glycoxidation" ([@B3]) were introduced at an early stage in this research field to highlight the importance of oxidation chemistry in AGE formation. Reactive oxygen species (ROS) and free (decompartmentalized) metal ions were identified as key participants in the Maillard reaction, and chelators were identified as potent inhibitors of browning and cross-linking of proteins by glucose. Oxygen was described as a fixative of irreversible damage to proteins via the Maillard reaction, and today metal-catalyzed oxidation reactions and chemical modifications of proteins, including numerous AGEs ([Fig. 1](#F1){ref-type="fig"}), advanced lipoxidation end products (ALEs), and protein oxidation products, are implicated in many chronic diseases involving oxidative stress, including diabetes and cardiovascular and neurodegenerative diseases ([@B1]--[@B5]).
{#F1}
AGE inhibitors {#s3}
==============
Aminoguanidine. {#s4}
---------------
In 1986, Brownlee et al. ([@B6]) introduced the first AGE inhibitor, aminoguanidine, as a trap or scavenger of reactive carbonyl intermediates in the Maillard reaction ([Fig. 1](#F1){ref-type="fig"}). In numerous studies in animal models of both type 1 and type 2 diabetes, aminoguanidine inhibited AGE formation in concert with inhibition of diabetic renal, retinal, neural, and vascular complications ([@B7]). Aminoguanidine is administered at a relatively high dose (typically 1 g/L in drinking water); in severely hyperglycemic rodents, which may consume their body weight in drinking water per day, this dose is equivalent to ∼1 g/kg/day. While the dose is enormous, it is not unreasonable; aminoguanidine has a short plasma half-life (∼1 h), and AGE inhibitors must be present at a concentration sufficient to continuously react with and trap chemical intermediates in the Maillard reaction ([Fig. 1](#F1){ref-type="fig"}). High aminoguanidine concentrations are required to drive sluggish and thermodynamically unfavorable trapping reactions to completion. More reactive carbonyl traps are likely to be toxic, e.g., because of their reaction with and depletion of vitamin B~6~, pyridoxal. While aminoguanidine is the prototype AGE inhibitor, its proposed mechanism of action is based completely on model chemical studies in vitro. Today, \>25 years since its discovery, there is no published evidence that aminoguanidine traps AGE precursors in vivo; i.e., none of the types of adducts described in [Fig. 1](#F1){ref-type="fig"} have been detected in urine or plasma.
Pyridoxamine. {#s5}
-------------
The B~6~ vitamer pyridoxamine was described as an Amadorin or post-Amadori AGE inhibitor, trapping products derived from the Amadori compound fructoselysine, the first stable glucose adduct to protein ([@B8]). Pyridoxamine is now considered to have multiple mechanisms of action: *1*) blocking oxidation of the Amadori intermediate; *2*) trapping of reactive carbonyl and dicarbonyl compounds derived from the Amadori compound; *3*) chelation of metal ion catalysts of oxidation chemistry; and *4*) scavenging of ROS ([@B9]). As observed with aminoguanidine, pyridoxamine inhibited the full range of diabetic vascular complications in animal models of both type 1 and type 2 diabetes ([@B8],[@B9]). In addition to their AGE-inhibitory activity, both aminoguanidine and pyridoxamine significantly lowered plasma triglycerides and cholesterol ([@B10]). Although pyridoxamine adducts of lipid peroxidation products have been identified in animal models of diabetes ([@B11]), like aminoguanidine, not a single product of reaction of pyridoxamine with a dicarbonyl intermediate in AGE formation, such as methylglyoxal or deoxyglucosones, has been identified in biological systems. Considering the sensitivity and specificity of modern mass spectrometric techniques, it seems unlikely that this is the result of technical limitations.
Other carbonyl-trapping agents. {#s6}
-------------------------------
A number of other compounds with reactive nucleophilic functional groups and carbonyl-trapping activity in vitro are effective as AGE inhibitors in rodent models of diabetes, including 2,3-diaminophenazine, OPB-9195, tenilsetam, penicillamine, and several derivatives of aminoguanidine ([Fig. 2*A*](#F2){ref-type="fig"}) ([@B4],[@B5],[@B12]). However, like aminoguanidine and pyridoxamine, no products of reaction of these compounds with intermediates in glycation or glycoxidation reactions have been identified in vivo. Although penicillamine, carnosine, and possibly lipoic acid have carbonyl-trapping activity in vitro, their potent chelating activity, measured by inhibition of metal-catalyzed ascorbate oxidation ([@B13]) ([Table 1](#T1){ref-type="table"}), would contribute to their AGE-inhibitory activity. To the best of our knowledge, analogs of these compounds with AGE-inhibitory activity, but lacking in chelating activity, have not been identified.
######
Structures of AGE inhibitors. *A*: First-generation AGE inhibitors. *B*: Novel LR compounds (adapted from Rahbar and Figarola \[[@B18]\]).
######
IC~50~ of various compounds that inhibit metal-catalyzed oxidation of ascorbate\*
Thiamine and benfotiamine. {#s7}
--------------------------
Thiamine and benfotiamine are often described as AGE inhibitors but have a much broader spectrum of action. In a unifying hypothesis on the mechanism of diabetes complications ([@B14],[@B15]), Brownlee ([@B14]) and Giacco and Brownlee ([@B15]) proposed that inhibition of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) by mitochondrial-derived ROS is a pivotal step in the pathogenesis of diabetes complications. The resultant accumulation of glycolytic intermediates leads to activation of the polyol, hexosamine, and diacylglycerol--protein kinase C pathways and increased formation of methylglyoxal-derived AGEs and expression of the receptor for AGE (RAGE). Thiamine and benfotiamine correct these metabolic changes, it is proposed, by diverting glucose metabolism from glycolysis through activation of transketolase, the thiamine-dependent enzyme in the pentose phosphate pathway (PPP) ([@B16]). Indeed, benfotiamine reverses all of the observed metabolic changes induced by hyperglycemia in vitro, including formation of methylglyoxal-derived AGEs; it also inhibits complications in diabetic animals and has shown promise in clinical trials ([@B16]). Transgenic overexpression of superoxide dismutase, which would limit ROS-dependent inactivation of GAPDH, also reversed many of the metabolic changes, as well as AGE formation, and inhibited development of complications in diabetic mice ([@B15]).
Studies with thiamine and benfotiamine illustrate that alterations in enzymatic pathways in diabetes may have a significant impact on oxidative stress and formation of AGEs. Glyceraldehyde-3-phophate, a major precursor of methylglyoxal, is one product increased as a result of inhibition of GAPDH by mitochondrial-derived ROS. Glyceraldehyde-3-phophate is not just a source of methylglyoxal AGEs but is also among the most readily autoxidized sugars and thereby a source of superoxide for metal-catalyzed Fenton and Haber-Weiss reactions, leading to increased production of hydroxyl radicals. Effects on the concentration of hexose phosphates would explain reported effects of thiamine and benfotiamine on formation of other AGEs. However, benfotiamine also has anti-inflammatory effects in diseases other than diabetes ([@B16]), e.g., endotoxemia, suggesting that its mechanism of action is not limited to activation of the PPP or other thiamine-dependent pathways. There are, to our knowledge, no reports on the chelating activity of thiamine or benfotiamine, but both thiamine, in the form of thiamine diphosphate, and benfotiamine, because of its complex multidentate structure, are likely to be effective in chelation of metal ions.
Lalezari-Rahbar compounds and the effects of chelation. {#s8}
-------------------------------------------------------
Lalezari-Rahbar (LR) compounds (named after the developers) are a new class of AGE inhibitors, a diverse group of aromatic organic acids with ureido and carboxamide functional groups ([Fig. 2*B*](#F2){ref-type="fig"}) ([@B17],[@B18]). These compounds have been shown to protect against complications in animal models of diabetes and to have profound lipid-lowering effects ([@B19],[@B20]), mimicking the effects of aminoguanidine and pyridoxamine in diabetic rats ([@B10]). The common effects of the AGE inhibitors on AGEs and plasma lipids suggest similar underlying mechanisms of action. Yet, the LR compounds lack nucleophilic groups and do not trap carbonyl compounds under physiological conditions, even in vitro. Regardless, they are potent inhibitors of AGE formation in diabetic rats and mice at doses of 50 mg/L in drinking water, i.e., at ≤5% of the molar dose commonly used for aminoguanidine and pyridoxamine. The LR compounds appear to inhibit AGE formation primarily through their chelating activity; they are potent inhibitors of metal-catalyzed ascorbate oxidation, with half-maximal inhibition (IC~50~) observed in the micromolar (micromole per liter) range ([Table 1](#T1){ref-type="table"}). Another recently described class of AGE inhibitors, derivatives of edaravone, has also been shown to have poor carbonyl-trapping activity but potent AGE-inhibitory activity ([@B21]). The IC~50~ for inhibition of ascorbate oxidation by TM2002, the most effective of these compounds, was 84 μmol/L ([Table 1](#T1){ref-type="table"}).
The LR compounds and edaravone derivatives have provided novel insight into the mechanism of action of AGE inhibitors. They suggest that chelation alone, independent of carbonyl trapping, is sufficient to inhibit AGE formation and protect against diabetes complications. Chelators would not only inhibit autoxidative glycosylation and glycoxidation but also inhibit enzymatic and metal-catalyzed ROS production after ligation of AGEs with scavenger receptors, such as RAGE. Thus, LR-90 inhibits the inflammatory response (increased expression of RAGE, monocyte chemoattractant protein-1, cyclooxygenase-2, and NADPH oxidase) in monocytes stimulated with the RAGE ligand, S100b, and also blocks the increase in monocyte endothelial cell adhesion ([@B22]). LR-90 also inhibited nuclear factor-κB activation after activation of the inflammatory response by tumor necrosis factor-α, suggesting that AGE inhibitors may exert wide-ranging effects on oxidative stress and inflammation by mechanisms other than carbonyl trapping.
Chelation is a common, but commonly overlooked, characteristic of most drugs with multiple functional groups. The chelating activity of AGE inhibitors varies widely: aminoguanidine and pyridoxamine inhibit ascorbate oxidation with IC~50~ in the 1--5 mmol/L range, while carnosine, diaminophenazine, OPB-9195, and tenilsetam are effective in the 5--50 μmol/L range ([Table 1](#T1){ref-type="table"}) ([@B23]). Though rigorous comparisons have not been conducted, the weaker the chelator, e.g., aminoguanidine or pyridoxamine, the higher the concentration generally used to inhibit the Maillard reaction in vitro and, in general, the higher the dose required to achieve efficacy in vivo. Even a weak chelator, however, in sufficient amount may deplete free or weakly bound metal ions by promoting their excretion in urine or bile.
ACE inhibitors, angiotensin receptor blockers, and other antihypertensive agents {#s9}
================================================================================
AGE inhibitors are not the only drugs that inhibit AGE formation. Miyata et al. ([@B24],[@B25]) demonstrated that ACE inhibitors (ACEIs) and angiotensin receptor blockers (ARBs) inhibit the formation of AGEs in Maillard reactions in vitro. These compounds were poor carbonyl traps and were proposed to act primarily by inhibiting the oxidative formation, rather than by trapping, of reactive carbonyl intermediates. Several of the ACEIs and ARBs were shown to be potent inhibitors of metal-catalyzed oxidation of ascorbate, with IC~50~ generally in the 1--10 μmol/L range ([Table 1](#T1){ref-type="table"}). Forbes et al. ([@B26]) demonstrated that the ACEI ramipril, at ∼10 μmol/L in drinking water (∼0.1% the dose of aminoguanidine), was comparable with aminoguanidine in inhibiting renal AGE formation and albuminuria in diabetic rats. In these studies, renal nitrotyrosine was also decreased in both ramipril- and aminoguanidine-treated animals, suggesting that both compounds inhibit AGE formation, in part, through effects on ROS production. Liu et al. ([@B27]) showed that renoprotection by the ACEI benazepril in a spontaneously hypertensive (nondiabetic) rat model was also accompanied by a decrease in AGE accumulation and expression of RAGE in the kidney, as well as decreased expression of NADPH oxidase p47phox and other biomarkers of oxidative stress. Monacelli et al. ([@B28]) reported that the ARB valsartan also decreased plasma and urinary AGEs in type 2 diabetic patients.
The relationship between the AGE-inhibitory and chelating activities of ACEIs and ARBs has not been systematically studied. The IC~50~ for ascorbate oxidation is 110 μmol/L for the ACEI temocaprilat ([@B24]) and \<10 μmol/L for AVE8048, the active metabolite of the combination ACEI/vasopeptidase inhibitor AVE7688, which was a more potent AGE inhibitor ([@B29]). Izuhara et al. ([@B30]) evaluated more than 100 ARB derivatives and focused on the compound R147176, which had potent AGE-inhibitory activity in vitro and also a low (∼3 μmol/L) IC~50~ for ascorbate oxidation. Although R147176 had low affinity for the angiotensin receptor 1 and had minimal blood pressure--lowering activity, it was an effective inhibitor of nephropathy in three different rat models of hypertension and/or diabetes. R147176 has a core structure common to several ARBs, suggesting that the benefits of many ARBs are at least partly the result of their chelating and antioxidant activity ([Table 1](#T1){ref-type="table"}). Nangaku et al. ([@B31]) reported that the antihypertensive agent hydralazine, which does not interact with the renin-angiotensin system, also protected against nephropathy in the diabetic rat and was a potent AGE inhibitor both in vitro and in vivo; hydralazine has an IC~50~ for ascorbate oxidation ∼1 μmol/L ([@B31]). In their comparison of the ARB olmesartan (IC~50~ ∼5 μmol/L) with hydralazine, the authors noted peak plasma concentration of ∼10 μmol/L for both drugs in clinical studies, indicating that tissue concentrations of the drugs should be sufficient to inhibit metal-catalyzed oxidation and glycoxidation reactions in vivo.
Miyata et al. ([@B32]) have summarized the evidence that antihypertensive agents have a range of renoprotective effects, independent of their blood pressure--lowering activity. Prominent among these effects is inhibition of oxidative stress, resulting in *1*) decreased AGE formation, *2*) decreased renal iron deposition, *3*) decreased RAGE expression, and *4*) decreased infiltration of inflammatory cells. All of these observations are consistent with the chelating and antioxidant activity of antihypertensive agents, which would limit both formation of AGEs and activation of AGE-RAGE--mediated inflammatory signaling cascades. The fact that ACEIs and ARBs are so effective in inhibition of AGE formation may explain in part why AGE inhibitors, despite their introduction over 25 years ago ([@B6]), have not found use in clinical management of diabetes complications: they may provide little added protection in patients already being treated with ACEIs and ARBs.
AGE breakers {#s10}
============
The original AGE breaker, *N*-phenacylthiazolium bromide (PTB) ([@B33]), its dimethylthiazolium analog alagebrium (ALT)-711 ([@B34]), and recently described pyridinium analogs TRC4186 and TRC4149 ([@B35],[@B36]) ([Fig. 3](#F3){ref-type="fig"}) were designed to cleave AGE cross-links in tissue proteins. As support for their mechanism of action, these compounds *1*) released AGE albumin from preformed AGE-albumin-collagen complexes, *2*) released immunoglobulins bound to red cells of diabetic rats, and *3*) reversed or decreased collagen cross-linking in diabetic rats. Despite these observations, not a single AGE cross-link structure identified in tissue proteins to date ([Fig. 4](#F4){ref-type="fig"}) contains a dicarbonyl structure susceptible to the proposed mechanism of action of AGE breakers ([Fig. 3](#F3){ref-type="fig"}). Indeed, the proposed target of AGE breakers, dicarbonyl cross-links, would be reactive carbonyl compounds, which are unlikely to accumulate in protein. Regardless of their target or mechanism of action, AGE breakers increase vascular elasticity and improve cardiovascular function in aging humans, in animal models of diabetes, and in other diseases ([@B5],[@B34],[@B35],[@B37]).
![Structures (*top panel*) and proposed mechanism of action (*bottom panel*) of AGE breakers. According to this scheme, the dinucleophilic AGE breaker adds across a dicarbonyl AGE cross-link, followed by an internal rearrangement to cleave the dicarbonyl bond, breaking the cross-link. Following hydrolysis, the AGE breaker is regenerated, leaving a chemically inert CML on one peptide and a chemically reactive aldehyde functional group on the other peptide involved in the cross-link. (Adapted from Vasan et al. \[[@B33]\].)](549fig3){#F3}
{#F4}
Yang et al. ([@B38]) demonstrated that although AGE breakers cleaved dicarbonyl structures in model compounds, they did not cleave cross-links in insoluble skin or tendon collagen of diabetic rats or cleave cross-links in RNase polymerized by reaction with glucose in vitro. In all of the studies demonstrating the activity of AGE breakers in vitro, fresh rat skin or tail collagens (unprocessed by dialysis or acid extraction to remove labile intermediates or cross-links) were used for analysis of cross-link breaking activity. In contrast, the lack of AGE-breaking activity was demonstrated with the acetic acid--extracted insoluble fraction of skin collagen, which would lack the labile (reversible) intermediates and cross-links. Similarly, red cells may have proteins bound to their surface membranes by labile Schiff base (imine) bonds. These, like the labile Schiff base fraction of glycated hemoglobin, would dissociate during incubation with AGE breakers. Overall, it is unlikely that AGE breakers react according to the reaction scheme shown in [Fig. 3](#F3){ref-type="fig"}. Indeed, one of the proposed products of AGE breaking is the AGE N^ε^-(carboxymethyl)lysine (CML), which has been identified as a ligand for RAGE and would thereby initiate a proinflammatory response. In addition, the protein is left with a four-carbon carbohydrate fragment with a reactive aldehyde group, which would participate in further AGE formation. Thus, it is difficult to rationalize beneficial effects from the proposed mechanism of action of AGE breakers.
Despite the lack of evidence for their AGE-breaking activity, AGE breakers are potent chelators, and their hydrolysis products have even stronger chelating activity than the intact compounds ([@B23]) ([Table 1](#T1){ref-type="table"}). In addition, all of the activities ascribed to AGE breakers in in vitro assays, including dissociation of AGE-albumin complexes and immunoglobulin adducts on red cells, and in vivo effects on collagen cross-linking have been replicated in detail with several LR compounds ([@B17],[@B18]), which lack functional groups that could cleave dicarbonyl compounds. In summary, despite several animal model and clinical studies that convincingly demonstrate the merits of AGE breakers for treatment of cardiovascular and renal pathology in diabetes and aging ([@B4],[@B5],[@B34]--[@B37]), the effects of these compounds are unlikely to be related to their proposed mechanism of action and may be explained largely, if not completely, by their chelating and antioxidant activities. Chelation, by inhibition of glycoxidation reactions, would limit the progression from AGE precursor to AGE cross-link, enabling the rejuvenation of the extracellular matrix by turnover of cross-linked proteins and biosynthesis of native matrix proteins.
Aldose reductase inhibitors {#s11}
===========================
Aldose reductase inhibitors (ARIs) are designed to block excessive metabolism of glucose through the sorbitol pathway, thereby protecting against accumulation of fructose, dicarbonyl intermediates, and imbalances in NADPH- and glutathione-dependent antioxidant defenses. ARIs have a diverse range of structures, with side effects often attributable to their lack of inhibitor specificity ([@B39]). While their chelating activity has not been investigated systematically, all of them contain functional groups with potential chelating activity, including carboxyl, amino, imino, imidazole, oxo, thio, and hydroxyl groups. Wolff and colleagues showed that ARIs inhibited metal-catalyzed oxidation of ascorbate and lipids ([@B40]), and Nakamura et al. ([@B41]) reported that the protective effects of the ARI NZ-314, a trioxoimidazolidine derivative, against diabetic neuropathy were in large part mimicked by the copper chelator trientine (triethylenetetramine). There are also several reports that ARIs inhibit AGE formation in diabetes and aging ([@B5],[@B42]). Although they may act by reducing the concentration of AGE precursors, secondary mechanisms of action, such as chelation, may be important in understanding the broader impacts of these compounds on the pathology of diabetes and aging.
Chelators {#s12}
=========
Alterations in iron and copper homeostasis are a characteristic feature of diabetes, evidenced by deposition of iron and copper in heart, kidney, and other tissues ([@B43],[@B44]). Underlying the alterations in copper homeostasis are defects in copper absorption, tissue distribution, and an increase in loosely bound (chelatable) copper in tissues ([@B43]--[@B45]). Cooper (summarized in [@B45]) reported that, compared with control subjects, type 2 diabetic patients had higher basal urinary copper excretion and a greater increase in urinary copper in response to oral administration of triethylenetetramine, a well-tolerated copper chelator used clinically for treatment of Wilson disease. Administration of triethylenetetramine for 6 months caused a significant ∼5% reversion of left ventricular mass toward normal in type 2 diabetic patients compared with an increase of 3% in left ventricular mass in untreated patients; these changes occurred without effects on blood pressure or blood glucose concentration. Like AGE inhibitors and breakers, triethylenetetramine also protected rats with streptozotocin (STZ)-induced diabetes against development of cardiovascular and renal pathology, including left ventricular dysfunction and collagen accumulation, renal fibrosis, and albuminuria. Triethylenetetramine also affected the renal proteome, altering the expression of numerous proteins involved in cellular metabolism, ion transport, and oxidative stress in the kidney of rats with STZ-induced diabetes. In later studies, Baynes and Murray ([@B46]) demonstrated that triethylenetetramine inhibited structural and functional changes in the heart and kidney of Zucker diabetic rats, a model of type 2 diabetes. Consistent with the studies in type 1 animal models and clinical studies ([@B46]), protection of cardiac function in type 2 diabetic rats was achieved without effect on blood pressure or blood glucose concentration. The effect of triethylenetetramine on tissue AGEs is still unknown.
Baynes and Murray ([@B46]) and Nagai et al. ([@B47]) evaluated the effects of citrate, a relatively nonspecific chelator, on progression of diabetes complications. Citrate, at the same dose as triethylenetetramine (1 g/L in drinking water), provided comparable protection against cardiac structural and functional changes in the Zucker type 2 diabetic rat. In rats with STZ-induced diabetes, citrate also inhibited albuminuria, cataractogenesis, and ketosis ([@B47]). The effects of citrate on ketosis have previously been reported, but the study of Nagai et al. ([@B47]) demonstrated that citrate also inhibited formation of two AGEs, both CML and N^ε^-(carboxyethyl)lysine (CEL), in the lens. This is the first evidence that even a common dietary chelator found in citrus fruits and drinks may inhibit AGE formation and protect against development of diabetes complications. Although citrate is known to promote intestinal absorption of copper and iron, there is no information on its overall effects on metal ion homeostasis in vivo or on AGE formation in other tissues. There are numerous reports in the literature on the benefits of plant extracts and nutraceuticals (e.g., rutin, polyphenols, quercetin, resveratrol) against diabetes complications. Some of these compounds also inhibit AGE formation in diabetic animals, and it is possible that they may work, in part, by limiting the uptake or promoting the excretion of metal ions through chelating activity.
One of the puzzling aspects of the studies is that triethylenetetramine, which is a weak iron chelator and does not appear to affect overall iron balance ([@B45]), has such an impact on diabetes complications, when both iron and copper excess are apparent in diabetes ([@B43],[@B48]). It is also clear that Wilson disease, which is characterized by severe copper overload in tissues and high levels of chelatable copper, does not lead to the characteristic complications of diabetes, although this might be obscured by the severe hepatic disease. Although iron is strongly implicated in the pathogenesis of diabetes and its complications and although specific iron chelators, such as desferoxamine, have shown beneficial effects in treatment of diabetes both in animal models and in clinical studies ([@B48],[@B49]), particularly in patients with iron overload diseases, the iron-chelating activity of various drugs has not been well studied, possibly because of the facile oxidation and precipitation of iron salts at neutral pH under air. It is likely that the copper and iron work together in diabetes; the disorder in copper homeostasis may lead to oxidation of iron and deposition of insoluble, but potentially redox active, ferric iron in tissues. These metals could cocatalyze oxidative stress and resultant inflammation, probably exacerbated by mitochondrial dysfunction and hypoxia in diabetes. The rapid elimination of decompartmentalized copper by triethylenetetramine would intercept the ROS-generating cycle, possibly leading to a more gradual leaching of iron from tissues and then to restoration of vascular tone.
Conclusions and perspective {#s13}
===========================
We have presented evidence here that the activity of AGE inhibitors and breakers on the formation of AGEs and development of diabetes complications may be explained in large part by their chelating activity or the chelating activity of products of their hydrolysis and/or metabolism. In addition, we propose that the AGE-inhibitory activity of other drugs commonly used for treatment of diabetes complications, including ACEIs, ARBs and ARIs, may also be attributed to their chelation activity. These compounds, as well as AGE inhibitors and breakers, have a diverse range of structures, and other functions clearly come into play with specific compounds, e.g., antihypertensive agents; however, chelation stands out as a likely, common mechanism of their action on AGE formation. Higher doses are required for weaker and more polar chelators, such as aminoguanidine and pyridoxamine, while more hydrophobic compounds, such as the LR compounds, ACEIs, ARBs, and ARIs, work effectively at lower doses, probably because of differences in bioavailability, plasma, or tissue half-lives or metabolism. Hydrophobic chelators may also penetrate and concentrate in membranous compartments and remove loosely bound metal ions from subcellular compartments. It is also possible that some chelators may shift the redox potential of iron or copper, affecting their catalytic activity in a way that could exacerbate oxidative stress and diabetes complications. To date, there are few reports on the effects of any of these compounds on metal-ion homeostasis, and AGEs have not been measured in renal or vascular tissues or in collagen after treatment with chelators. Such experiments are essential to test the hypothesis presented in this article that chelation is a fundamental mechanism of action of the wide range of compounds that affect AGE formation in diabetes. While we have focused this discussion on AGEs and AGE inhibitors and breakers, lipoxidation reactions proceed in concert with glycoxidation and oxidation in tissues and AGE inhibitors also inhibit lipoxidation reactions ([@B11]), so it will be important to determine the extent to which the formation of ALEs, as well as other biomarkers of oxidative stress in diabetes, is inhibited by chelators. Finally, the interplay between iron and copper in oxidative stress, diabetes, and its complications deserves greater attention. Both iron and copper chelators inhibit AGE formation in model systems in vitro, and it is equally likely that these metals have both unique and complementary roles in vivo.
In conclusion, although "Quackwatch" appears among the first three URLs in a Bing, Google, or Yahoo search for "chelation therapy," we argue that there is good reason to invest in further research on chronic, low-dose, oral chelation therapy for treatment of diabetes and its complications. This recommendation also applies to cardiovascular and neurodegenerative diseases ([@B50]) and other diseases with a chronic inflammatory component, marked by the accumulation of iron and copper, AGEs and ALEs, and protein and DNA oxidation products at sites of pathology. Chelation should inhibit the metal-catalyzed oxidation reactions and inflammatory processes leading to and resulting from accumulation of oxidative damage in tissues.
Research in the authors' laboratories was supported by a Grant-in-Aid for Scientific Research (18790619) from the Ministry of Education, Japan (to R.N.), the University of South Carolina School of Medicine Center for Biomedical Research Excellence (to D.B.M.), and research grants DK071283 (to T.O.M.) and DK19971 (to J.W.B.) from the National Institute of Diabetes and Digestive and Kidney Diseases. Pacific Northwest National Laboratory is supported by the U.S. Department of Energy (DOE) Office of Biological and Environmental Research and is operated by Battelle for the DOE under contract no. DE-AC06-76-RLO-1830. J.W.B. receives royalties from NephroGenex, Inc., through a licensing agreement with the University of South Carolina for use of pyridoxamine. No other potential conflicts of interest relevant to this article were reported.
R.N., D.B.M., and T.O.M. contributed to discussion, suggested additional topics for consideration and evaluation, and reviewed and edited the manuscript. J.W.B. wrote the original draft of the manuscript. J.W.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
The authors thank Dr. Thomas K. Borg (Medical University of South Carolina), Dr. Garth J.S. Cooper (Manchester University, Manchester, U.K.), Dr. Alicia J. Jenkins (University of Melbourne, Melbourne, Australia), and Dr. Kevin J. Wells-Knecht (Cephalon, Inc.) for constructive comments on the manuscript.
[^1]: R.N. and D.B.M. contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijms-21-03946}
===============
Matrix metalloproteinases (MMPs) are a family of zinc-dependent endoproteases responsible for tissue remodeling and degradation of extracellular matrix (ECM) proteins \[[@B1-ijms-21-03946]\]. MMPs are secreted by endothelial cells, vascular smooth muscle, fibroblasts, osteoblasts, macrophages, neutrophils, and lymphocytes \[[@B2-ijms-21-03946]\]. MMPs family contain 28 members, 23 are expressed in human tissues and 14 are expressed in veins and arteries. They are classified on the basis of their substrates and the organization of their structural domains \[[@B3-ijms-21-03946]\]. MMPs may be inhibited by tissue and biological and synthetic inhibitors \[[@B4-ijms-21-03946]\]. Endogenous tissue inhibitors of metalloproteinases (TIMPs) are widely distributed in many tissues and organs. In the case of biological inhibitors, doxycycline (an antibiotic) remains the single MMPs inhibitor that was approved by Food and Drug Administration (FDA) for clinical use \[[@B5-ijms-21-03946]\]. Importantly, statins exert pleiotropic effects in vivo, including influence on signaling mechanisms that leads to MMPs inhibition. \[[@B6-ijms-21-03946]\]. MMPs play important role in maintaining vein wall structure and function, but on the other hand, in adverse cardiovascular remodeling, atherosclerotic plaque formation and plaque instability \[[@B7-ijms-21-03946],[@B8-ijms-21-03946]\]. Serum elevation of MMP-2, ADAMTS-1, and ADAMTS-7 is correlated with the initial stages of chronic venous disease (CVD), whereas the serum elevation of MMP-1, MMP-8, MMP-9, NGAL, ADAM-10, ADAM-17, and ADAMTS-4 is particularly involved in skin change complications \[[@B9-ijms-21-03946]\].
Increased activity of MMP-7 and MMP-9 was observed in unstable plaques, the highest tissue expression of MMP-9 was found in plaques of lipid type compared with plaques of necrotic and inflammatory-erosive types \[[@B10-ijms-21-03946]\]. In addition, MMP-7 could contribute to plaque instability in carotid atherosclerosis, potentially involving macrophage-related mechanisms \[[@B11-ijms-21-03946]\]. MMP-9 is a strong independent predictor of atherosclerotic plaque instability in stable coronary heart disease (CHD) patients, where MMP-9 levels are positively associated with the size of the necrotic core of coronary atherosclerotic plaques \[[@B12-ijms-21-03946]\]. It was shown that serum MMP-9 and the MMP-9/TIMP-1 molar ratio may be valuable in acute coronary syndrome (ACS) diagnosis and prognosis. MMP-9 activation in serum was associated with poor cardiovascular outcome \[[@B13-ijms-21-03946]\]. Tan et al. investigated associations of MMP-9 and monocyte chemoattractant protein-1 (MCP-1) concentrations with the severity of carotid atherosclerosis, based on measurements of carotid plaque and intima--media thickness (IMT). Elevated serum MMP-9 concentration was independently associated with high total carotid artery plaque score, plaque instability, and large IMT value \[[@B14-ijms-21-03946]\]. Importantly, using selective MMP-9 inhibitors may provide new perspective to intervene on ECM remodeling in humans \[[@B15-ijms-21-03946]\]. These data confirm that MMPs may be biomarkers and have been proposed as potential therapeutic targets in cardiovascular diseases \[[@B16-ijms-21-03946],[@B17-ijms-21-03946],[@B18-ijms-21-03946]\].
2. MMPs Structure and Tissue Distribution {#sec2-ijms-21-03946}
=========================================
Matrix metalloproteinases (MMPs) are Zn^2+^ endopeptidases. They contain about 80 types of amino acids, a catalytic metalloproteinase domain (containing about 170 amino acids), a binding type of the special peptide or hinge-shaped region of variable length and a hemopexin stationary domain, with about 200 amino acids \[[@B19-ijms-21-03946]\]. A definition of MMPs is based on their substrate specificity, structural organization, and cellular location. From a chemical point of view, the MMPs include collagenases (MMP-1/collagenase-1, MMP-8/collagenase-2, and MMP-13/collagenase-3), stromelysins (MMP-3/stromelysin-1, MMP-10/stromelysin-2, and MMP-11/stromelysin-3), gelatinases (MMP-2/gelatinase-A and MMP-9/gelatinase-B), membrane-type MMPs such as transmembrane MMP-14, MMP-15, MMP-16, MMP-24, membrane-anchored MMP-17, and MMP-25, and matrilysins (MMP-7/matrilysin-1 and MMP-26/matrilysin-2 \[[@B20-ijms-21-03946]\] ([Figure 1](#ijms-21-03946-f001){ref-type="fig"}).
The MMP-1, MMP-2, MMP-3, MMP-7, and MMP-9 expression was found in endothelial cells and vascular smooth muscle cells (VSMCs), while MMP-12 showed expression in VSMCs and fibroblasts \[[@B21-ijms-21-03946]\]. MMP-1, MMP-3, MMP-7, MMP-9, MMP-13, a membrane-type (MT) MMPs, MT-MMP1, and MT-MMP3 were found in the vascular wall \[[@B22-ijms-21-03946]\]. Indeed leukocytes and dermal fibroblasts are key sources of MMP-2, whereas platelets are source of MMP-1, MMP-2, MMP-3, and MMP-14 \[[@B23-ijms-21-03946]\] ([Table 1](#ijms-21-03946-t001){ref-type="table"}).
3. Extracellular Vesicles as MMPs Carriers {#sec3-ijms-21-03946}
==========================================
Extracellular vesicles (EVs) as structures secreted by cells by the paracrine route include exosomes (diameter range: 30--100 nm), apoptotic bodies (diameter range: 0.1--1 μm), microvesicles (MVs, diameter range: \> 1 μm), and large oncosomes (diameter range: 1--10 μm). They are released from cells as a result of fusion of the endosome with the plasma membrane. EVs participate in intercellular communication \[[@B24-ijms-21-03946],[@B25-ijms-21-03946],[@B26-ijms-21-03946]\]. The composition of the microvesicles depends on the cell that secretes these structures. It is determined by a set of sorting proteins \[[@B27-ijms-21-03946]\]. The mechanisms of transport of MMPs to extracellular vesicles are not fully understood. Sinha's and Clark's researches indicated that cortactin, as a regulator of matrix metalloproteinase secretion, was involved in this process \[[@B28-ijms-21-03946],[@B29-ijms-21-03946]\]. In addition, the role of Rabi GTPases and kinesins in MMPs exposure to stem cell surface and packaging of these proteins into exosomes has been demonstrated \[[@B30-ijms-21-03946],[@B31-ijms-21-03946]\]. In turn, the location of MMP-14 in exosomes depends on vesicle-associated membrane protein 3 (VAMP-3) \[[@B32-ijms-21-03946]\].
EVs are involved in the secretion of MMPs into the intercellular space. As carriers of numerous proteins, including matrix metalloproteinases, extracellular matrix metalloproteinase inducer (EMMPRIN), and tissue inhibitors of metalloproteinases (TIMPs), they can affect the reconstruction of the intercellular matrix, resulting in, among others atherosclerotic plaques destabilization \[[@B33-ijms-21-03946]\]. Moreover, it has been shown that EVs not only transport MMPs but also participate in their activation. Studies by Bobryshev et al. have shown that at preatherosclerotic disease stage, the secretion of MMP-enriched EVs from arterial smooth muscle cells is significantly greater in the athero-prone areas of the human aorta than in the athero-resistant areas \[[@B34-ijms-21-03946]\]. Research by Laghezza et al. showed the involvement of EVs in the proteolytic activation and secretion of MMP-9. Extracellular vesicles secreted by endothelial cells and platelets, as MMP-9 carriers, participate in the regulation of atherosclerotic plaque neovascularization. Angiogenesis is promoted at low levels of endothelial-derived microvesicles containing active MMP-2 and MMP-9, while high levels of these MVs inhibit plaque angiogenesis which reduces the risk of rupture \[[@B35-ijms-21-03946]\]. In addition, Lozito et al. demonstrated that EVs secreted from endothelial cells are involved in the activation of MMP-2 initiating remodeling of the vascular matrix \[[@B36-ijms-21-03946],[@B37-ijms-21-03946]\]. MMP-2 present in EC-derived vesicles also participates in neovascularization, promoting capillary structure formation and plaque rupture \[[@B25-ijms-21-03946],[@B38-ijms-21-03946]\]. It is known that MT1-MMP present in exosomes secreted by fibroblasts participates in the pro-MMP2 activation process \[[@B39-ijms-21-03946]\]. In turn, studies in ApoE (-/-) mice have showed that olmesartan significantly limit the formation of atherosclerotic lesions in the aorta by reducing the activity of MMP-2 \[[@B40-ijms-21-03946]\].
MMP-rich microvesicles are effective modulators of weakening of the fibrous cap. Neutrophil-derived MVs enriched in MMP-9, metalloproteinase domain-containing proteins 10 (ADAM-10), and -17 (ADAM-17), as well as matrix-derived MVs rich in MMP-2 and endothelial MMP-10-enriched MVs, play a huge role in this process \[[@B41-ijms-21-03946]\]. In addition, exosomes secreted by various cells, including macrophages, transporting MMP-2, MMP-3, MMP-9, and MMP-13, simultaneously contribute to vascular calcification. This increases plaque susceptibility to rupture, leading to life-threatening cardiovascular events such as myocardial infarction \[[@B35-ijms-21-03946],[@B42-ijms-21-03946],[@B43-ijms-21-03946],[@B44-ijms-21-03946]\]. Studies in patients with chronic kidney disease have shown that under mineral imbalance, there is an increased secretion of MMP-2-rich exosomes from vascular smooth muscle cells, which converts the exosomes into loci to endothelial calcification \[[@B45-ijms-21-03946]\]. Chondrocyte-derived MVs containing MMP-3 participate in the calcification process of extracellular vesicles in the fibrous collagen of the extracellular matrix of cardiovascular tissues. Their secretion is induced by phosphate through extracellular kinases regulated by the Erk1/2 signal pathway \[[@B33-ijms-21-03946]\].
Exosomes may serve as biomarkers for the development of atherosclerosis, providing potential roles for diagnosis and treatment \[[@B46-ijms-21-03946]\]. Quantitative analysis of MMP-rich extracellular vesicles (especially those containing MMP-2 and -9) can be used as a parameter to assess calcification and plaque neovascularization in monitoring the clinical course of cardiovascular diseases. MVs analysis also seems to be helpful in identifying patients at risk of cardiovascular disease. MVs can also be a therapeutic target for preventing and controlling the development of atherosclerosis, including progression of plaque formation and its instability \[[@B40-ijms-21-03946],[@B47-ijms-21-03946]\].
4. Biological Inhibitors of MMPs {#sec4-ijms-21-03946}
================================
Among tissue inhibitors of MMPs (TIMPs), such as tetracyclines, chemically-synthesized small-molecular weight MMP inhibitors and inhibitory antibodies, only doxycycline, which was approved by FDA, has been evaluated extensively in patients \[[@B48-ijms-21-03946]\]. In a TIPTOP study, doxycycline decreased infarction size, and showed improvement of cardiac contractility characteristic in patients with acute ST-segment elevation myocardial infarction (STEMI) and left ventricular (LV) dysfunction \[[@B49-ijms-21-03946]\]. Additionally, in another clinical trial, short-term treatment with doxycycline caused reduction of MMP-2 activity in coronary bypass patients \[[@B50-ijms-21-03946]\]. It was shown that doxycycline attenuates heart mechanical dysfunction and reduces inflammation in patients with coronary artery disease (CAD) \[[@B51-ijms-21-03946]\]. In addition to its potential plaque stabilization in acute coronary syndromes properties, doxycycline shows promise in preventing ischemia-reperfusion injury and left ventricular remodeling \[[@B52-ijms-21-03946]\]. Furthermore, clinical study demonstrated that doxycycline therapy led to downregulation of MMPs, dampening cardiac inflammation, and reduction of secondary myocardial infarction (MI) risk in coronary bypass patients \[[@B53-ijms-21-03946]\].
The statins may inhibit expression of MMPs in atherosclerotic plaques by reducing vascular inflammation \[[@B54-ijms-21-03946],[@B55-ijms-21-03946]\]. Atorvastatin inhibits MMP-1, MMP-2, and MMP-9 expression in human retinal pigment epithelial cells \[[@B56-ijms-21-03946]\] and MMP-1, MMP-2, MMP-3, and MMP-9 secretion from rabbit macrophages and cultured rabbit aortic and human saphenous vein VSMCs \[[@B57-ijms-21-03946]\]. Komukai et al. in the EASY-FIT clinical trial indicated that atorvastatin increased fibrous cap thickness and plaque stability and decreased levels of MMP-9 in CAD patients \[[@B58-ijms-21-03946]\]. Further, Liu et al. confirmed that atorvastatin reduced plasma MMP-9 levels and myocardial dysfunction in STEMI \[[@B59-ijms-21-03946]\]. Li et al. demonstrated that simvastatin significantly suppressed LPS-induced MMP-9 release and mRNA expression in a time- and concentration-dependent manner \[[@B60-ijms-21-03946]\]. It was shown that in a rat model of heart failure, pravastatin suppressed the increase in myocardial MMP-2 and MMP-9 activity \[[@B61-ijms-21-03946]\]. In addition, rosuvastatin inhibits the expression of MMP-2 and MMP-9 \[[@B62-ijms-21-03946]\]. Moreover, a recent study reported the efficiency of combined therapy with rosuvastatin and ezetimibe to treat plaque instability and cardiovascular inflammation in CAD patients, by a significant decrease in MMP-9 plasma concentration \[[@B63-ijms-21-03946]\]. Sapienza et al. recently showed that patients with advanced PAD, a pathological condition of hemodynamic instability of the atheromatous plaque, had good results from the association between platelet aggregation inhibitors and statins in terms of patency of the reconstructions \[[@B64-ijms-21-03946],[@B65-ijms-21-03946]\].
5. MMPs in Pathogenesis of Atherosclerosis {#sec5-ijms-21-03946}
==========================================
Atherosclerosis is a chronic inflammatory disease of the vessel wall that is largely driven by an innate immune response \[[@B66-ijms-21-03946]\]. In pathogenesis of atherosclerosis, a pivotal role is played by innate immunity receptors such as toll-like receptors (TLR) and receptors for advanced glycation end products (RAGE) \[[@B67-ijms-21-03946]\]. TLR and RAGE mediate in macrophages and leukocyte recruitment and are significantly involved in the initiation and progression of atherosclerosis \[[@B68-ijms-21-03946]\]. This process is characterized by the accumulation of lipids, smooth muscle cell proliferation, cell apoptosis, necrosis, and fibrosis \[[@B69-ijms-21-03946]\]. MMPs plays a key role in all stages of atherosclerosis through vascular inflammation, endothelial dysfunction, smooth muscle cell migration, vascular calcification, extracellular matrix degradation, and plaque activation and destabilization \[[@B8-ijms-21-03946],[@B16-ijms-21-03946],[@B70-ijms-21-03946]\]. The expression of MMPs is controlled by different microRNA molecules. MicroRNA regulation of extracellular matrix components is crucial in the process of atherosclerotic plaque destabilization \[[@B71-ijms-21-03946]\]. In the artery wall, subendothelial retention of plasma lipoproteins triggers monocyte-derived macrophages and T helper type 1 (Th1) cells to form atherosclerotic plaques. Plaque rupture or endothelial erosion can induce thrombus formation, leading to myocardial infarction or ischemic stroke \[[@B72-ijms-21-03946]\]. Ruptured plaques are characterized by a large lipid-rich core, a thin fibrous cap that contains few smooth muscle cells and many macrophages, angiogenesis, adventitial inflammation, and remodeling. Plaque rupture is the most common cause of coronary thrombosis \[[@B73-ijms-21-03946]\]. It was shown that endothelial-to-mesenchymal transition (EndMT) is a cellular reprogramming mechanism by which endothelial cells acquire a mesenchymal phenotype. EndMT contributes to the initiation and progression of atherosclerotic lesion and plaque destabilization, which further causes acute cardiovascular events \[[@B74-ijms-21-03946]\].
5.1. MMPs in Vascular Inflammation and Recruitment of Immune Cells {#sec5dot1-ijms-21-03946}
------------------------------------------------------------------
MMPs participate in the immune response and play a key role in vascular inflammation that is strongly associated with atherosclerosis \[[@B75-ijms-21-03946]\]. Exposure of vascular wall cells to inflammatory mediators produced under chronic inflammation may lead to excessive MMPs activity in the arterial wall resident and recruited cells \[[@B70-ijms-21-03946],[@B76-ijms-21-03946]\]. In response to proinflammatory mediators, monocytes enhance MMPs expression and play a key role in inflammatory cell migration and invasion into the arterial wall \[[@B77-ijms-21-03946]\]. Human monocytes not only express TIMP-1, TIMP-2, MMP-8, MMP-12, and MMP-19 but also activate MAP kinase and NF-κB, and in this way promote the expression of MMP-2, MMP-3, MMP-10, and MMP-14. The ratio between MMPs and their inhibitors (TIMPs) to free active MMPs is critical in pathogenesis of cardiovascular diseases \[[@B78-ijms-21-03946]\]. TIMP-2 plays a protective role in the pathogenesis of atherosclerosis by suppressing MMP-14-dependent monocyte/macrophage accumulation into plaques \[[@B79-ijms-21-03946]\]. MMP-1 expression is augmented in inflammatory and autoimmune diseases through upregulation by inflammatory cytokines such as TNF α (tumor necrosis factor-α) and IL-1 (interleukin-1) \[[@B80-ijms-21-03946]\]. These inflammatory cytokines participate in ROS (reactive oxygen species) production and influence the expression and activity of MMPs. In addition, activation of the PDGFR-β and ERK1/2 atherogenic pathways participates in the production of MMP-1 in human coronary smooth muscle cells (SMCs) stimulated by oxidized low-density lipoproteins (oxLDL) \[[@B81-ijms-21-03946]\].
5.2. MMPs in Lipid Accumulation {#sec5dot2-ijms-21-03946}
-------------------------------
Migration of macrophages as well as deposition of low-density lipoprotein (LDL) under the endothelium initiate the process of atherosclerosis \[[@B82-ijms-21-03946]\]. Oxidatively modified LDL (oxLDL) are involved in the transformation of macrophages into foam cells and the migration of VSMCs into the intima \[[@B83-ijms-21-03946]\]. Foam cell macrophages, VSMCs, and endothelial cells produce cytokines enhancing the recruitment of inflammatory cells, which in turn attracts VSMCs into the neointima \[[@B48-ijms-21-03946]\]. MMPs and TIMPs are secreted constitutively or after activation by inflammatory response not only by monocytes and macrophages but also by foam cells \[[@B84-ijms-21-03946]\]. oxLDL activates MMP-2 through upregulation of the MT1-MMP expression and stimulation of oxidative radicals production by xanthine/xanthine oxidase \[[@B85-ijms-21-03946]\]. In slow-to-heal wounds increased activity and expression of MMP-1, MMP-8, and MMP-9 have been shown, accompanied with low TIMPs levels \[[@B86-ijms-21-03946]\]. It was shown that MMP-9 can modulate cholesterol metabolism, through MMP-9-plasma secreted phospholipase A~2~ axis \[[@B87-ijms-21-03946]\].
5.3. MMPs in Endothelial Dysfunction {#sec5dot3-ijms-21-03946}
------------------------------------
MMPs have been found to be involved in vascular wall remodeling and atherosclerosis development through inflammatory activation and endothelium dysfunction \[[@B88-ijms-21-03946]\]. The dysfunction of endothelium is characterized by proinflammatory and prothrombic state; the action of the endothelium is shifted toward reduced vasodilation, and the phenotype changes from antiadhesive to proadhesive \[[@B89-ijms-21-03946]\]. Damage of endothelial junctions results in enhanced endothelial permeability, which facilitates infiltration of various inflammatory mediators. Activation of endothelial MMP-2 can induce endothelial dysfunction and disintegrity \[[@B90-ijms-21-03946]\]. Recruited vascular wall cells can remodel the surrounding extracellular matrix through MMPs that affect migration, proliferation, and apoptosis of endothelial cells and VSMCs \[[@B91-ijms-21-03946]\].
5.4. MMPs in Migration of VSMCs {#sec5dot4-ijms-21-03946}
-------------------------------
MMPs contribute to intimal thickening and vessel wall remodeling in atherosclerosis by promoting migration and proliferation of VSMCs \[[@B92-ijms-21-03946],[@B93-ijms-21-03946]\]. MMPs degrade ECM and facilitate VSMCs migration into the intima. Studies have shown that MMP-2 is involved in these alterations of VSMCs behaviors. IL-1β, cytokine responsible for VSMCs migration, activates MMP-2 synthesis \[[@B94-ijms-21-03946]\]. MMP-2 plays a key role in ox-LDL-induced activating VSMCs proliferation through various pathways \[[@B95-ijms-21-03946]\]. TNF-α promotes VSMCs migration through MMP-9 upregulation \[[@B96-ijms-21-03946]\]. Enhanced expression of MMP-9 in VSMCs can be caused by angiotensin II (Ang II) that participates in atherosclerosis by NF-κB pathways and angiotensin type 1 receptor \[[@B97-ijms-21-03946]\]. Moreover, MMP-9 stimulates proliferation of VSMCs by regulating cell adhesion and cadherin association \[[@B98-ijms-21-03946]\]. Increased serum MMP-10 levels were connected with increased carotid intima--media thickness, atherosclerotic plaques, and inflammatory markers in patients with preclinical atherosclerosis \[[@B99-ijms-21-03946]\]. Activity of certain variant of the MMP-3 promoter was correlated with progression of luminal narrowing of coronary arteries \[[@B100-ijms-21-03946],[@B101-ijms-21-03946]\].
5.5. MMPs in Plaque Neovascularization {#sec5dot5-ijms-21-03946}
--------------------------------------
Remodeling of extracellular matrix and vascular basement membrane allow to form new blood vessels. Studies have shown that MMP-9 mobilizes vascular endothelial growth factor (VEGF) from the ECM, which can contribute to plaque neovascularization. The increase of VEGF bioavailability results from the ECM proteolysis. \[[@B102-ijms-21-03946]\]. Hypoxia and inflammation in the lesion can induce plaque neovascularization, which leads to plaque rupture \[[@B103-ijms-21-03946]\]. In hypoxia, the direct target of injury are endothelial cells. Hypoxia induces cell proliferation in the vascular wall and results in increased MMP-9 expression in VSMCs \[[@B104-ijms-21-03946]\]. These cells enhance MMPs activity in hypoxic areas. The interaction of macrophages with VSMCs influences neovascularization and destabilization of plaque because of enhanced MMP-1 and MMP-9 production \[[@B105-ijms-21-03946]\]. In neovascularization, MMPs play a crucial role because of their participation in penetration of the ECM and tissue remodeling. The cross-talk between VSMCs and macrophages augments synthesis of angiogenic factors. Both cells produce not only VEGF and IL-1β (cytokine, stimulating the secretion of VEGF) but also MMPs responsible for degradation of ECM components, resulting in penetration and reshaping of connective tissue.
5.6. MMPs in Plaque Calcification {#sec5dot6-ijms-21-03946}
---------------------------------
MMPs were also demonstrated to provoke vascular calcification \[[@B106-ijms-21-03946]\]. Vascular calcification is determined as an attribute of advanced atherosclerotic plaques. The plasma level of MMP-7 is positively associated with carotid calcification \[[@B107-ijms-21-03946]\]. Additionally, MMP-2 contributes to the calcification of SMCs, participating in the formation of vascular calcified lesions. MMP-2 deficiency in ApoE-/- mice related to suppressed calcium deposition in aorta-derived cultured SMCs \[[@B108-ijms-21-03946]\]. In ApoE-/- mice with advanced atherosclerotic lesion, chondrocyte-like cells were found and several bone-related proteins were expressed \[[@B109-ijms-21-03946]\]. Platelets contain and release MMPs within vascular injury \[[@B110-ijms-21-03946]\]. MMP-2 participates in thrombogenesis, cleaving platelet proteinase-activated receptor 1 (PAR1), and stimulating the aggregation of platelets \[[@B111-ijms-21-03946]\]. Purroy et al. research showed low MMP-10 expression, while significantly reducing the area of atherosclerosis and their calcification in ApoE-/- Mmp10 -/- mice. Furthermore, studies in subjects with subclinical atherosclerosis showed a correlation between MMP-10 serum activity and the degree of coronary calcification \[[@B112-ijms-21-03946]\].
5.7. MMPs in Plaque Activation and Destabilization {#sec5dot7-ijms-21-03946}
--------------------------------------------------
Studies indicate the involvement of MMPs in atherosclerotic plaque initiation, progression, and plaque rupture \[[@B113-ijms-21-03946]\]. Link between inflammation and plaque vulnerability might provide increased CRP-related MMPs activation \[[@B114-ijms-21-03946]\]. MMP-1 may contribute to the progression of the human atherosclerotic lesions by remodeling of the neointimal extracellular matrix (ECM) \[[@B115-ijms-21-03946]\]. An important role in the pathogenesis of atherosclerosis and plaque rupture may be played by enhanced MMP-1 expression induced by monocyte--endothelial cell interactions \[[@B116-ijms-21-03946]\].
Advanced atherosclerotic plaque is filled by cell debris, extracellular lipids, and foam cell macrophages, whereas stable plaque is protected from the circulating blood by a fibrous cap. Plaque destabilization can be caused by proteolysis of fibrillar collagens. MMP-1, MMP-8, and MMP-13 show this collagenolytic activity in unstable plaques \[[@B117-ijms-21-03946],[@B118-ijms-21-03946]\]. Studies on human cells infected with influenza A virus showed an increase of MMP-13 expression, which may partly explain the destabilization of atherosclerotic plaques. Analysis of the corresponding changes in the ApoE-deficient mouse model has shown that the increase of MMP-13 expression is due to activation of the p38 mitogen-activated protein kinase signaling pathway \[[@B119-ijms-21-03946]\]. Increased MMP-7 activity can be connected with the apoptosis of VSMCs in the fibrous cap and lesion destabilization \[[@B120-ijms-21-03946]\]. Enhanced MMP-7 activity is involved in the cleavage of N-cadherin, which mediates VSMCs apoptosis, resulting in plaque instability \[[@B121-ijms-21-03946]\]. Vulnerable atherosclerotic plaques are characterized by increased MMPs levels \[[@B122-ijms-21-03946]\]. Activated VSMCs produce high-mobility group box 1 (HMGB1), which affects the expression of MMP-2, MMP-3, and MMP-9 and enhances plaque rupture \[[@B123-ijms-21-03946]\] ([Figure 2](#ijms-21-03946-f002){ref-type="fig"}).
Genetic variants of MMP-9 were found to be associated with different stages of atherosclerosis \[[@B124-ijms-21-03946]\]. Crucial significance of MMP-9 in the growth of atherosclerotic plaque is shown in ApoE-deficient mice \[[@B125-ijms-21-03946]\]. Atherosclerotic plaque formation is associated with hypertension-induced atherosclerosis and increased levels of MMP-9 mRNA \[[@B126-ijms-21-03946]\]. Matrix-degrading activity of MMPs contributes to complications of human atherosclerotic lesions. Studies have shown enhanced expression of MMPs in macrophage-rich areas (around lipid-rich core) suggesting important role of MMPs (derived from macrophages) in developing of vulnerable regions of atherosclerotic plaques \[[@B127-ijms-21-03946]\]. Ablation of MMP-9 contributed to the reduction of atherosclerotic burden and attenuated macrophage infiltration and collagen deposition. In addition, imbalance in the expression of MMP-9/TIMP-1 was perceived in macrophages within the atherosclerotic plaques \[[@B128-ijms-21-03946]\]. A positive correlation between MMP-9 macrophages and plasma lysophosphatidic acid (LPA) concentration was also shown. This is particularly important due to the fact that MMP-9 is derived from macrophages with coronary plaque instability. As demonstrated by Gu et al., LPA acting through the LPA2 receptor can induce MMP-9 by activating the NF-κB pathway \[[@B129-ijms-21-03946]\] ([Table 2](#ijms-21-03946-t002){ref-type="table"}).
Evidence suggests that neutrophil gelatinase-associated lipocalin (NGAL) may play a crucial role in vascular remodeling and plaque instability during the development of atherosclerosis \[[@B130-ijms-21-03946]\]. NGAL is involved in the regulation of MMP activity. MMPs and NGAL also play a key role in development of arterial aneurysms \[[@B131-ijms-21-03946]\]. The complex formation between NGAL and MMP-9 suggests that NGAL might play a role in progression of atherothrombotic disease. NGAL has performed as a more consistent predictor of adverse outcomes in patients with ACS, especially STEMI, compared with patients having stable CAD \[[@B132-ijms-21-03946]\]. NGAL was associated with increased risk of long-term CHD events, independent of conventional risk factors and biomarkers \[[@B133-ijms-21-03946]\].
6. MMPs as Biomarkers in Cardiovascular Diseases {#sec6-ijms-21-03946}
================================================
MMPs are involved in various stages of plaque progression that are important to predict future atherothrombotic cardiovascular events \[[@B134-ijms-21-03946]\]. Vulnerable atherosclerotic plaques are responsible for life-threatening clinical endpoints; thus, the best approach for both stroke and myocardial infarction (MI) would be safe diagnosis and knock out of the vulnerable lesion before endpoints occur \[[@B135-ijms-21-03946]\]. Higher markers of matrix remodeling, such as MMP-9 and TIMP-1, were associated with a greater prevalence of carotid stenosis and subclinical atherosclerosis in the internal carotid artery (IC) \[[@B136-ijms-21-03946]\]. Sequence variants at the MMP-1 genomic locus may influence risk of coronary heart disease in humans \[[@B137-ijms-21-03946]\]. MMP-1 (-1607 1G/2G) and MMP-3 (-1171 5A/6A) polymorphisms may contribute to different subtypes of ischemic stroke susceptibility \[[@B138-ijms-21-03946]\].
Since 2005, Sapienza and coauthors showed that an imbalance exists between MMPs and TIMPs in unstable carotid plaques, which is reflected in the plasma levels of these markers \[[@B139-ijms-21-03946]\]. Disturbed equilibrium of the metalloproteinase/tissue inhibitors system, destabilization of atherosclerotic plaque, and acute coronary syndrome (ACS) in patients are caused by increased expression of MMP-2 and MMP-9 metalloproteinases and their tissue inhibitor (TIMP-2) \[[@B140-ijms-21-03946]\]. Studies have indicated that MMP-2, MMP-8, and MMP-9 facilitate plaque rupture and clinical events \[[@B18-ijms-21-03946],[@B118-ijms-21-03946]\]. In 2006, Turu et al. demonstrated significantly higher MMP-8 plasma concentrations in patients with unstable atherosclerotic plaques compared to patients with stable plaques \[[@B141-ijms-21-03946]\]. The results of recent research confirm the relationship between MMP-8 and the processes of remodeling and destabilization of the plaque. A demonstrated relationship between the variability of the MMP-8 gene and atherosclerosis suggest that MMP-8 concentrations may have prognostic and diagnostic significance in assessing the patient's cardiovascular risk \[[@B142-ijms-21-03946]\]. MMP-8 has also been shown to be a proteinase responsible for the activation of other MMPs, such as MMP-2 and MMP-9 \[[@B143-ijms-21-03946],[@B144-ijms-21-03946]\]. MMP-2 promotes a platelet aggregation, which is involved in a prothrombotic effect of atherosclerotic plaques. Enhanced MMP-2 activity was especially observed in plaques of patients with a higher rate of subsequent major cardiovascular events \[[@B145-ijms-21-03946]\]. MMP-2 has been determined as a predictor of mortality in patients with ACS \[[@B146-ijms-21-03946]\]. High levels of MMP-8 in the carotid plaque correlate with an unstable plaque composition and systemic cardiovascular outcome \[[@B147-ijms-21-03946]\]. Seifert et al. confirmed that MMP-2 and MMP-9 activities was significantly higher in the ApoE-/- cuff model in unstable atherosclerotic plaques as compared to downstream more stable plaque phenotypes \[[@B148-ijms-21-03946]\]. Increased plasma MMP-9 and TIMP-1 levels have been demonstrated in the coronary circulation in patients with acute coronary syndrome (ACS), which suggests active process of plaque rupture and future risk of cardiovascular events \[[@B149-ijms-21-03946]\]. Kelly et al. have shown the association of MMP-9 and TIMP-1 with function and remodeling of left ventricular (LV) as well as cardiovascular death or heart failure \[[@B150-ijms-21-03946]\]. The correlation between TIMP-1 and MMP-9 and echocardiographic parameters of LV remodeling after AMI (acute myocardial infarction) suggest a promising indicator of patients with adverse prognosis. In type 2 diabetes mellitus, elevated MMP-7 and MMP-12 plasma levels are linked with severe atherosclerosis as well as more frequent coronary events \[[@B151-ijms-21-03946]\]. The stronger elevated plasma levels of MMP12 and imbalance between MMP12 and TIMP1 was observed in patients with STEMI compared to patients with stable angina pectoris \[[@B152-ijms-21-03946]\]. Moreover, another study has shown elevated peripheral blood levels of MMP-2 and MMP-9 in patients with ACS \[[@B153-ijms-21-03946]\]. Blankenberg et al. demonstrated that plasma MMP-9 concentrations may constitute risk markers for future cardiovascular death in ACS patients \[[@B154-ijms-21-03946]\]. However, this study also has shown the correlation between MMP-9 level and acute-phase reactants; therefore, independent prognostic information of MMPs is not clear. Moreover, increased levels of TIMPs are associated with enhanced risk of cardiovascular events in patients with ACS \[[@B155-ijms-21-03946],[@B156-ijms-21-03946]\]. Elevated MMPs levels have been shown not only in affected tissue in patients with ACS, arthritis, and cancer but also in peripheral blood, suggesting elevated blood MMPs levels in patients at risk of acute plaque disruption \[[@B153-ijms-21-03946]\]. Higher plasma MMP-9 levels have been shown in patients with significant carotid stenosis undergoing carotid endarterectomy with detected spontaneous embolization, compared to patients without it \[[@B157-ijms-21-03946]\]. MMP-9 expression and macrophage infiltration constitute strong indicators of high risk atherosclerotic carotid plaques and plaque instability \[[@B158-ijms-21-03946],[@B159-ijms-21-03946]\]. MMP-9 may have vital significance in differentiating patients with unstable angina with or without plaques \[[@B160-ijms-21-03946]\]. In patients with angiographically confirmed coronary heart disease (CHD), it was showed that increased concentrations of MMP-9 at baseline were associated with future cardiovascular (CV) death \[[@B161-ijms-21-03946]\]. Tang et al. demonstrated that increased plasma levels of MMP2 and MMP9 in patients with coronary heart diseases (CHD) suggest the instability of the atherosclerotic plaque in correlation to the severity of ACS, and may serve as good indicators for the prediction of ACS and diagnosis of chronic total occlusion (CTO) of the coronary artery \[[@B162-ijms-21-03946]\]. MMP-2 and MMP-9 circulating levels can serve as indicators of efficiency of the therapy provided to heart failure (HF) patients, as well as for identification of patients who could benefit from particular therapeutic intervention via modification of MMP pathway \[[@B163-ijms-21-03946]\] ([Figure 3](#ijms-21-03946-f003){ref-type="fig"}).
7. Conclusions {#sec7-ijms-21-03946}
==============
Accumulating evidence confirm the key role of MMPs in plaque development and pathogenesis of atherosclerosis, especially in the advanced stages of the disease, where their elevated activity increases the risk of plaque rupture. The degradation of extracellular matrix protein is catalyzed by MMPs and participates in the migration of vascular SMCs and consequently, leads to plaque instability. Increased peripheral blood MMP-2 and MMP-9 in acute coronary syndrome (ACS) may be useful as noninvasive tests for detection of plaque vulnerability. Local carotid plaque MMP-8 level corresponds with a higher frequency of secondary manifestations of cardiovascular disease during follow-up, and increased plasma levels are predictive for subsequent all-cause mortality. Mentioned findings suggest that local MMPs plaque levels may be useful in detection of the vulnerable plaque and can help to predict patients with high risk of atherosclerotic cardiovascular events. The use of biomarkers to select patients for individualized therapies in secondary prevention will help achieve the goal of precision medicine. \[[@B164-ijms-21-03946]\]. In addition, modulation of immune-mediated inflammation is a new promising point of action for the eradication of fatal atherosclerotic events \[[@B135-ijms-21-03946]\]. The Canakinumab Anti-Inflammatory Thrombosis Outcome Study (CANTOS) trial confirmed that targeting inflammation with interleukin-1β (IL-1 β) inhibition significantly reduces cardiovascular events \[[@B165-ijms-21-03946]\].
This research received no external funding.
The authors declare no conflict of interest.
{#ijms-21-03946-f001}
{#ijms-21-03946-f002}
{#ijms-21-03946-f003}
ijms-21-03946-t001_Table 1
######
The family of matrix metalloproteinases (MMPs), tissue distribution, and substrates.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SUBFAMILY MMP Tissue Distribution Substrates
------------------------ -------- --------------------------------------------------------------------------------- ---------------------------------------------------------------------------------
**Collagenases**
Collagenase-1 MMP-1 Endothelial cells, VSMCs, vascular wall, platelets, fibroblasts, macrophages Collagens: I, II, III, VII, VIII, X\
Gelatins, entactin, aggrecan, link protein
Collagenase-2 MMP-8 Macrophages, neutrophils Collagens: I, II, III;\
Aggrecan link protein
Collagenase-3 MMP-13 Vascular wall, SMCs, macrophages Collagens: I, II, III
**Gelatinases**
Gelatinase-A MMP-2 Endothelial cells, VSMCs, adventitia, leukocytes, dermal fibroblasts, platelets Collagens: I, IV, V, VII, X, XI;\
Gelatins, elastin, fibronectin, laminin, b-amyloid protein precursor
Gelatinase-B MMP-9 Endothelial cells, VSMCs, adventitia, vascular wall, macrophages Collagens: IV, V, XIV;\
Gelatins, elastin, entactin, vitronectin
**Stromelysins**
Stromelysin-1 MMP-3 Endothelial cells, VSMCs, vascular wall, platelets Collagens: III, IV, IX, X;\
Aggrecan, fibronectin, laminin, elastin, gelatins, casein
Stromelysin-2 MMP-10 Uterus Collagens: II, IV, V;\
Aggrecan, fibronectin, gelatins, activate procollagenase
Stromelysin-3 MMP-11 Uterus, brain Collagen IV,\
weakly fibronectin, laminin, aggrecan, gelatins, IGFBP-1, a1-protease inhibitor
**Matrilysins**
Matrilysin-1 MMP-7 Endothelial cells, VSMCs, vascular wall, uterus Collagen IV,\
aggrecan fibronectin, laminin, entactin, vitronectin, casein, IGFBP-1
Matrilysin-2 MMP-26 Collagen IV,\
gelatin, fibronectin
**Membrane-type MMPs**
MT1-MMP MMP-14 Vascular wall, platelets, fibroblasts, uterus, brain Collagens I, II, III;\
fibronectin, laminin-1, dermatan sulfate
MT2-MMP MMP-15 Fibroblasts, leukocytes Large tenascin-C, fibronectin, laminin, entactin, aggrecan, perlecan
MT3-MMP MMP-16 Vascular wall, leukocytes Collagen III, gelatin, casein, fibronectin
MT4-MMP MMP-17 Brain Activates MMP2 by cleavage
MT5-MMP MMP-24 Leukocytes, lung, pancreas, kidney, brain Activates MMP2 by cleavage
MT6-MMP MMP-25 Leukocytes Inactivates alpha-1 proteinase inhibitor
**Other**
Metalloelastase MMP-12 VSMCs, fibroblasts, macrophages, great saphenous vein Elastin, fibronectin
RASI-1 MMP-19 Liver Gelatin
Enamelysin MMP-20 Tooth enamel Amelogenin
Xenopus-MMP MMP-21 Fibroblasts, macrophages, placenta
CA-MMP MMP-23 Ovary, testis Gelatin
Epilysin MMP-28 Skin, keratinocytes Casein
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
ijms-21-03946-t002_Table 2
######
Participation of MMPs in processes involved in development and progression of atherosclerotic plaque.
----------------------------------------------------------------------------- ------------------------------------------------------------------
**Vascular Inflammation and Endothelial Dysfunction**
Increase of inflammatory cell migration and invasion into the arterial wall MMP-1, MMP-8, MMP-9, MMP-12, MMP-13, MMP-14
Influence on endothelial dysfunction MMP-2, MMP-9
Participation in oxLDL effect MMP-1, MMP-2, MMP-8, MMP-9
Promotion of EC apoptosis MMP-9
**Development of Plaque**
Increase of intima--media thickness MMP-1, MMP-2, MMP-3, MMP-7, MMP-8, MMP-9, MMP-10
Promotion of plaque growth MMP-2
Promotion of VSCMs migration MMP-9
Decrease of VSCMs contractility MMP-2, MMP-9
**Neovascularization**
Degradation of ECM MMP-1, MMP-8, MMP-9, MMP-12, MMP-13, MMP-14
Release of growth factors MMP-2, MMP-9, MMP-13, MMP-14
**Calcification**
Provoke vascular calcification MMP-2, MMP-7
**Plaque Activation and Destabilization**
Apoptosis of VSMCs in the fibrous cap MMP-7
Collagenolytic activity MMP-1, MMP-8, MMP-13
Enhances plaque rupture MMP-1, MMP-2, MMP-3, MMP-7, MMP-8, MMP-9, MMP-12, MMP-13, MMP-14
----------------------------------------------------------------------------- ------------------------------------------------------------------
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
DNA methylation plays a crucial role in gene expression regulation, maintenance of genome stability, and it controls the transcription of invading and mobile DNA elements ([@B24]; [@B14]). Plants possess four types of DNA methyltransferases (MTases), namely methyltransferase (MET), chromomethylase (CMT), domains rearranged methyltransferase (DRM), and DNA methyltransferase homologue 2 (DNMT2) ([@B24]). MET maintains CG methylation of heterochromatic regions enriched with transposable elements (TEs) and repeats, and genic regions ([@B7]; [@B25]). CMT and DRM mediate CHG and CHH (H=A/C/T) methylation ([@B24]; [@B23]). DNMT2 has a novel transfer RNA (tRNA) methyltransferase activity ([@B19]; [@B21]), but its role in C5 DNA methylation remains largely unknown ([@B30]).
DNA methyltransferases genes have been found in many plant species, such as tobacco, rice, *Arabidopsis*, wheat, maize, *Physcomitrella*, and legumes ([@B9]; [@B41]; [@B30]; [@B16]; [@B36]; [@B27]; [@B32]). DNA methylation is primarily catalyzed by the DNA methyltransferase family. DNA methyltransferase plays an important role in plant development, transcriptional regulation, and metabolic pathway control. For example, the triple mutation of *drm1drm2cmt3* leads to delayed growth, small plant size, and partial barrenness in *Arabidopsis* ([@B3]). DNA methylation is also involved in tomato fruit ripening. The Colorless non-ripening (*Cnr*) mutation inhibits normal tomato ripening due to methylation of the *SBP-CNR* gene promoter ([@B28]; [@B17]). [@B4] recently reported on the role of a chromomethylase (SlCMT3) for the stable methylation of the promoter region of the *Cnr* gene.
Plants are continuously affected by abiotic or biotic environments, and thus have developed notable abilities to regulate their physiological and developmental mechanisms through gene expression regulation in response to these environmental perturbations ([@B44]). Epigenetic mechanisms, including DNA methylation and histone modification, play important roles in regulating gene expression in plant responses to environmental stress ([@B31]; [@B8]; [@B2]; [@B1]). For instance, salinity and water stress can trigger demethylation at coding regions of certain genes and subsequently initiate their expression ([@B6]). To the contrary, satellite sequences can be hypermethylated, especially in CHG sequences after salt stress ([@B11]). Low-temperature stress reduces the amount of methyltransferase in corn (*Zea mays L.*) ([@B37]).
In this study, based on the complete sequence of tomato genomes, as well as expression profiles at different tissues/stages and abiotic stresses (low temperature and salt), the nine tomato MTases were analyzed and characterized through an approach combining bioinformatics and expression experiments. Our study provides valuable information for functional research of DNA methyltransferase genes in tomato.
Materials and Methods
=====================
DNA and protein sequence analysis
---------------------------------
The protein sequences of *Arabidopsis* and rice MTases (Table S1) were used to search for the amino acid sequences of tomato MTases in the NCBI (<http://blast.ncbi.nlm.nih.gov/Blast.cgi>) and Sol Genomics Network (SGN) (<http://solgenomics.net/>) databases using the Blastp tool with the filter-off option and a cut-off value of 1 e^-10^. The genomic DNA sequences of these nine genes were obtained from the SGN. In order to analyze the exons and introns of genomic DNA, sequence alignment between CDS (coding sequence) and genomic DNA was done by MultAlin. The gene structures of the DNA MTases in tomato were generated using the GSDS. Molecular weight (Mw), isoelectric points, and grand average of hydropathicity (GRAVY) were estimated with the ExPASy compute Mw tool. Conserved structure domains were annotated based on ScanProsite and the Pfam protein family database. Motif detection was dependent on MEME ([@B40]). The phylogenetic tree was constructed using MEGA 5.02 software and the neighbor-joining method with the following parameters: bootstrap analysis of 1,000 replicates, Poisson model, and pairwise deletion. The numbers at the nodes indicate the bootstrap values. Promoter element analysis was performed using plant CARE and PLACE, which is a database of motifs found in plant *cis-*acting regulatory DNA elements.
Plant material
--------------
Tomato (*Solanum lycopersicum* Mill. cv. Ailsa Craig) seedlings were grown under greenhouse conditions (16 h days at 27 °C and 8 h nights at 19 °C). For organ-specific expression profiling of genes, tomato roots, stems, leaves, sepals, flowers and fruit pericarp tissues of different periods were harvested. Roots and stems were collected from 45-day-old tomato seedlings based on their uniformity. The leaves were taken from three different parts of 65-day-old tomato plants, namely young leaves (3 leaves of new growth), mature leaves (5 to 7 leaves from top to bottom) and senescent leaves (8 to 10 leaves from top to bottom). Sepals and petals were collected at the same time. Flowers were marked at anthesis and fruit development was recorded as days post-anthesis (DPA). Fruits ripening was divided into five stages, namely IMG (immature green, 28 DPA), MG (mature green, 35 DPA, full fruit expansion but no obvious color change), B (breaker, fruit showing the first signs of ripening-associated color change from green to yellow), B4 (4 days after breaker) and B7 (7 days after breaker).
Expression analysis of DNA MTase genes by gene microarray
---------------------------------------------------------
Microarray expression data were obtained from the tomato Gene Chip platform of Genevestigator (<https://www.genevestigator.com/gv/>). The nucleotide sequences of DNA MTase genes were used as query sequences to blast against all of the gene probe sequences from the Affymetrix Gene Chip (<http://www.affymetrix.com/>), and the best homologous probes were selected and used to carry out search in the Affymetrix Tomato Genome Array platform.
Stress treatments
-----------------
Potted 35-day-old tomato seedlings chosen based on their uniformity were used for all stress treatments. For salt stress treatment, the roots of tomato seedlings were submerged in a solution containing 250 mM NaCl for 0, 1, 2, 4, 8, 12, and 24 hours, and the young leaves of the treated seedlings and controls were collected. For low temperature stress treatment, the whole potted tomato seedlings were incubated at 4 °C for 0, 1, 2, 4, 8, 12, and 24 hours, after which the leaves were collected ([@B45]). All stress treatments were performed with three biological replicates.
RNA isolation and quantitative RT-PCR analysis
----------------------------------------------
Total RNA was extracted from tomato tissues with the Trizol reagent (Invitrogen, Shanghai, China). Genomic DNA pollution was eliminated with DNase I (Promega, Beijing, China) in the presence of RNase inhibitor (Takara Biotechnology, Japan). Poly (A)+RNA was used as a template for synthesis of first-strand cDNA. Complementary DNA was synthesized by M-MLV reverse transcriptase (Promega, Beijing, China) at 37 °C for 1 h. The quantitative RT-PCR reaction system and conditions were performed as in our previous report ([@B18]). The tomato *CAC* and *EF1*α genes were used as internal controls under normal growth conditions ([@B13]) and abiotic stress ([@B29]), respectively. The analysis of gene relative expression levels was conducted using the 2^-DDCT^ method ([@B26]). All primers used for quantitative RT-PCR are listed in Supplementary Table S2. The mean values of three independent experiments were calculated, and the standard deviations (± SD) were indicated.
Statistical analysis
--------------------
All experiments were conducted with three biological replicates. Statistical data were analyzed by Origin 8.0 software, and performed using the Student's *t*-test (SPSS 22.0). Values of *p* \< 0.05 were considered significant. Data are presented as mean ± SD.
Results
=======
Identification of tomato DNA MTases and sequence analysis
---------------------------------------------------------
Firstly, the data for 11 and 10 MTases in *Arabidopsis* and rice (Table S1) was collected from NCBI, respectively. Based on these data, nine MTases were identified in tomato through Blastp ([Table 1](#t1){ref-type="table"}). The open reading frame (ORF) length of these genes varies from 1.1 to 4.6 kb, and their protein length ranged from 381 to 1559 amino acids. All the deduced polypeptides are hydrophilic. In addition, [Figure 1](#f1){ref-type="fig"} shows the intron-exon organization (number of introns and exons) of nine MTases in tomato. The coding regions of CMT subfamily genes are interrupted by 14-21 introns ([Figure 1](#f1){ref-type="fig"}). MET gene (*SlMET1*) length is approximately 4.6 kb in tomato, harboring 12 exons. The length of the DRM subfamily genes in tomato varies from 1.8-2.1 kb with nine exons. DNMT2 gene (*SlMETL*) is smallest in length (1.1 kb) harboring nine exons. Genomic distribution of these tomato MTase genes was also analyzed. Nine tomato MTases genes are dispersedly located on chromosomes, with one MTase variant mostly located on a single chromosome ([Table 1](#t1){ref-type="table"}), suggesting at least partial influence of WGD in the diversification of the MTases family in tomato, rather than gene duplication.
{#f1}
###### Overview of MTases genes identified in tomato.
Gene name ORF length[^a^](#t1fn1){ref-type="table-fn"} (bp) Deduced polypeptide[^b^](#t1fn2){ref-type="table-fn"} Chromosome number Accession number[^c^](#t1fn3){ref-type="table-fn"}
----------- --------------------------------------------------- ------------------------------------------------------- ------------------- ---------------------------------------------------- -------- ------------------------ --------------
*SlMET1* 4680 1559 175.03 6.03 -0.517 ch11 18811587-18827974 AJ002140
*SlCMT2* 2802 933 104.50 5.40 -0.581 ch12 65430879-65437290 XM_004252792
*SlCMT3* 2235 808 91.17 4.90 -0.376 ch01 756827-764851 XM_004228549
*SlCMT4* 2667 888 100.04 8.82 -0.655 ch08:292101-303500 XR_182971
*SlDRM5* 1812 603 68.03 4.79 -0.504 ch02 29084337-29096121 EU344815
*SlDRM6* 1830 609 69.09 5.16 -0.464 ch10 59372041-59376567 SGN-U321564
*SlDRM7* 1824 607 68.71 4.75 -0.492 ch04 185839-189158 TC161581
*SlDRM8* 2100 699 78.82 5.45 -0.411 ch05:62542201-62559200 SGN-U325992
*SlMETL* 1146 381 43.42 5.44 -0.312 ch08 53192484-53203494 XP_004245195
Length of open reading frame in base pairs.
Length of amino acids, molecular weight (kDa), isoelectric point (pI), and grand average of hydropathicity (GRAVY) of the deduced polypeptide.
GenBank, SGN or TIGR accession number of tomato MTases genes.
Conserved domains and phylogenetic analysis
-------------------------------------------
Alignment of the amino acid sequences of these nine tomato DNA MTases revealed that tomato MTases genes possess a regulatory region and a catalytic region with conserved motifs that are arranged in a specific order. Six highly conserved motifs I, IV, VI, VIII, IX, and X were identified in the methyltransferase domain via MEME analysis in the nine MTases ([Figure 2](#f2){ref-type="fig"}). We found that each subfamily of tomato MTase has a characteristic arrangement of these motifs in the catalytic region. MET members showed the order of motifs as I, IV, VI, VIII, IX, and X. In CMT members, chromodomain was present between the conserved motifs I and IV with the rest of the arrangement similar to the MET members. It is interesting to note that *SlCMT4* appeared to lack the IX and X domains. DRM members showed the order of motifs as VI, VIII, IX, X, I, and IV except in *SlDRM7*, which only possesses the IV motif. Only one ubiquitin-associated domain (UBA) was present in the DRM family members. Similar to MET, DNMT2 member showed the order of motifs as I, IV, VI, VIII, IX, and X, but no regulatory region ([Figure 2](#f2){ref-type="fig"}).
{#f2}
MTases, including replication foci domain (RFD), bromo adjacent homology (BAH), and methyltransferase domains were classified as MET subfamily members, whereas members with the Chr domain, along with BAH, and methyltransferase domain were placed in the CMT subfamily ([Figure 2](#f2){ref-type="fig"}). Members harboring both UBA and methyltransferase domains were grouped into a DRM subfamily ([Figure 2](#f2){ref-type="fig"}). DNMT2 subfamily members seem to lack any amino-terminal regulatory domain and include only a methyltransferase domain ([Figure 2](#f2){ref-type="fig"}). In tomato, a total of three MTase genes were identified as CMT, one as MET, four as DRM, and one as DNMT2 members ([Figure 3](#f3){ref-type="fig"}); in *Arabidopsis*, three members belonged to CMT (AtCMT1, 2, and 3), four to MET (AtMET1, AtMET2a, AtMET2b, and AtMET3), three to DRM (AtDRM1, 2, and 3) and one to DNMT2 (AtDNMT2) families. Similarly, there were three CMTs (OsMET2a, OsMET2b, and OsMET2c), two METs (OsMET1-1 and OsMET1-2), four DRMs (OsDRM1aa, OsDRM1ba, OsDRM3, and OsZmet3) and one DNMT2 (OsDNMT2) in rice ([@B36]). As shown in [Figure 3](#f3){ref-type="fig"}, four clades (CMT, MET, DNMT2, and DRM) were clearly distinguished with support values close to 100. The CMT subfamily contained nine proteins, among which were three tomato proteins (SlCMT2, SlCMT3, and SlCMT4). The clades MET and DNMT2 included only SlMET and SlMETL, respectively. The DRM clade contained four tomato proteins (SlDRM5, SlDRM6, SlDRM7, and SlDRM8). Thus, our evolutionary analysis results showed good consistency with the classification results.
{#f3}
Transcription pattern of DNA MTase genes in wild-type tomato and mutants
------------------------------------------------------------------------
To elucidate the tissue/organ expression patterns of MTase genes in tomato, quantitative RT-PCR was carried out using cDNAs from different tissues and development stages. [Figure 4](#f4){ref-type="fig"} shows that *SlCMT2* was highly expressed in young leaves, mature green fruits, and stems, while its expression was down-regulated continuously during leaf development. *SlCMT3* was also predominantly expressed in young leaves and its transcription level declined continuously with further fruit ripening. *SlCMT4* was highly expressed in flowers and immature green fruits relative to other tissues, while its expression was down-regulated continuously during fruit development. The expression pattern of *SlMET1* was very similar to that of *SlDRM7*. Their transcripts both reached a maximum level in immature green fruits. *SlDRM5* was highly expressed in young leaves. During fruit development, *SlDRM5* transcripts reached a maximum in immature green fruit and then decreased. Interestingly, the expression of *SlDRM6* in the reproductive stage was higher than in the vegetative growth stage. *SlDRM8* expression was slightly higher in flowers, sepals, and immature green fruits than in other tissues. *SlMETL* expression was higher in ripening fruits and displayed an up-regulated tendency during fruit development. Spatial and temporal expression of *SlMET1*, *SlCMT2*, *SlDRM5*, *SlDRM7*, *SlDRM8*, and *SlMETL* were basically consistent with microarray expression data (Figure S1). Besides, it is worthy of note that the expression level of *SlMET1* in the tomato ripening mutants *rin* and *Nr* was significantly higher compared to wild-type tomato ([Figure 5](#f5){ref-type="fig"}).
{#f4}
{#f5}
Tomato DNA MTases are involved in abiotic stress response
---------------------------------------------------------
To further study the potential functions of these tomato DNA MTases genes, we carried out expression analyses under low temperature and salt stress conditions by quantitative RT-PCR. For low-temperature treatment ([Figure 6](#f6){ref-type="fig"}), we noted that the expression of *SlMET1* and *SlDRM5* was inhibited by low temperature and decreased gradually. The transcript levels of *SlCMT3*, *SlCMT4*, *SlDRM7*, *SlDRM8*, and *SlMETL* were also decreased under low temperature stress, especially *SlCMT3* and *SlDRM7*, which were sharply down-regulated at 1 h. Additionally, *SlCMT2* and *SlDRM6* were up-regulated slightly during the first 12 hours of treatment, but a significant decrease in *SlCMT2* mRNA was detected at 24 h.
{#f6}
For salt treatment ([Figure 7](#f7){ref-type="fig"}), the induction of *SlCMT2* gene expression was observed; it peaked at 4 h and returned to basal level at 24 h. The expression of *SlCMT3* in leaves was significantly up-regulated at 12 h by about 13-fold. *SlCMT4* was slightly down-regulated at 1 h and up-regulated subsequently in leaves. *SlDRM5* and *SlMETL* were induced, and their transcripts peaked at 4 h in leaves. The expression of *SlDRM6* was increased gradually and peaked at 4 h in leaves, with an expression pattern similar to that of *SlDRM7*. Comparatively, the transcript levels of *SlMET1* and *SlDRM8* were less affected in leaves. The above results suggest that these MTases genes may be involved in the response to salt stress.
{#f7}
Discussion
==========
DNA methylation is an important epigenetic modification established by DNA methyltransferase. Although tomato is a model plant for studying fleshy fruit development and ripening, little is known regarding a comprehensive analysis of MTases in tomato. In the present study, we analyzed tomato MTases and identified three members of CMT, one MET, four DRMs, and one DNMT2 in tomato. Each of the tomato MTases genes has a homologous gene in *Arabidopsis*, suggesting that MTases in tomato might have similar roles as in *Arabidopsis*. In addition, the systematic expression pattern of tomato MTases in different tissues/development stages and abiotic stress provides evidence for diverse functions in various aspects of plant development and abiotic stress responses.
The structural analysis suggested that catalytic DNA methylase domains are highly conserved, whereas the N-terminus, which is regarded as a regulatory region, is divergent ([Figure 2](#f2){ref-type="fig"}). Thus, these nine tomato MTase genes may play different roles in regulating tomato growth and development. MET subfamily members are very similar to the mammalian DNMT1 class ([@B24]). Our structural analysis of tomato CMTs (SlCMT2, SlCMT3, and SlCMT4) suggested that the N-terminus of CMT harbors the BAH and Chr domains, which could possibly enhance the binding attraction of CMTs to methylated histones, similar to *Zea mays* CMT3 ([@B10]). Four DRM members were identified in tomato. The N-terminus of DRM possesses the UBA domain, where sequence motifs occur that are usually involved in ubiquitin-mediated proteolysis and contributing to ubiquitin (Ub) binding or ubiquitin-like (UbL) domain binding. Recent findings have established DNMT2 as a tRNA methyltransferase that plays an important function under stress conditions ([@B34]; [@B39]). We also investigated one member (*SlMETL*) of the DNMT2 family in tomato, lacking a conserved N-terminal regulatory domain, but possessing a catalytic C-terminal domain, which seems to be characteristic for all DNMT2s.
So far, the characteristics and functions of MTases in *Arabidopsis* have been studied clearly ([@B15]), but there is very little knowledge of their expression profiles in different tissues/developmental stages in tomato ([@B38]). In this study, we investigated the expression pattern of the nine DNA MTases genes in different tissues/stages ([Figure 4](#f4){ref-type="fig"}), suggesting overlapping and specific roles during tomato development. The higher expression of *SlMET1* in IMG fruits in tomato suggested its role in the maintenance of methylation in early stages of fruit development. This is different from the expression of MET members in *Arabidopsis* and rice, which was higher in the early stages of flower and seed development ([@B33]; [@B42]; [@B22]; [@B36]; [@B35]). The ANAERO2CONSENSUS and CANBNNAPA elements ([@B12]) regulating fruit and embryo development respectively, were identified in the promoter of *SlMET1* (Table S3), suggesting *SlMET1* might be related with fruit development, which was confirmed by its high expression in fruit. *SlCMT4* was highly expressed in flower, immature green fruit, and young leaf, which was coincident with a previous report ([@B38]). *SlMETL* showed the highest expression in B4 fruits, and *SlDRM6* expression in reproductive stage was significantly higher than in vegetative growth stage, suggesting that these proteins may play an important role in tomato reproductive stage. Interestingly, *SlCMT3* was specifically expressed in young leaves, suggesting that *SlCMT3* may play critical roles in tomato leaf development. Consistent with its function in the DNA methylation maintenance, the tomato CMT was predominantly expressed in actively replicating cells in young leaves and roots. Additionally, it is noteworthy that *SlMET1* and *SlDRM7* were specifically expressed in immature green fruit, suggesting their useful application in fruit ripening and development.
Epigenetic modifications play an important role in response to environmental stimuli ([@B5]; [@B20]). For example, most of the MTases genes in pigeon pea are responsive to NaCl and extreme temperature ([@B32]). To further study the potential functions of the nine tomato MTases genes, we examined their expression under various stress conditions by quantitative RT-PCR. We found that most of the DNA MTases genes in tomato are responsive to stress treatments, including NaCl and low temperature ([Figures 6](#f6){ref-type="fig"} and [7](#f7){ref-type="fig"}), and the differential expression profiles indicated that they may function diversely in different stress conditions. Although *SlDRM5* and *SlDRM6* appeared highly similar in protein structure ([Figure 2](#f2){ref-type="fig"}) and transcription in native leaves ([Figure 4](#f4){ref-type="fig"}), the transcriptional responses to salt stress were remarkably different, being increased by about 2 times for *SlDRM5* and 4.5 times for *SlDRM6* after 4 h of treatment ([Figure 7](#f7){ref-type="fig"}). This probably correlates with number of GAAAAA (GT1GMSCAM4) promoter *cis-*elements, known to be responsible in wound repair (Table S3).
DNA methylation is involved widely in the regulation of the temporal and spatial gene expression in plants. DNA methyltransferase inhibitor 5-azacytidine induces tomato fruit premature ripening ([@B43]), and it is demonstrated that DNA methylation contributes to the regulation of fruit ripening. In this study, we observed that *SlMET1* was highly expressed in immature green fruit and then declined during fruit ripening, which was consistent with a previous report by [@B38]. Interestingly, the expression levels of *SlMET1* in the tomato ripening mutants *rin* and *Nr* are higher than in wild type tomato ([Figure 5](#f5){ref-type="fig"}), suggesting that *SlMET1* is negatively regulated by the ethylene signal and ripening-related transcriptional factor MADS-RIN. We speculate that the abnormal fruit ripening in the mutants *Nr* and *rin* might be related to the concurrent hypermethylation of multiple ripening-related genes by DNA methyltransferase SlMET1.
In summary, based on bioinformatics and transcriptional pattern analysis, the nine MTase genes identified in tomato could be involved in tomato development and abiotic stress responses. This study also provided valuable information about tomato MTase genes associated with fruit ripening.
This work was supported by the National Natural Science Foundation of China (no. 31801872), and by the scientific and technological project of DATONG SHANXI province (no. 2016046).
*Associate Editor: Marcia Pinheiro Margis*
Conflict of interest
====================
The authors declare that they have no conflict of interest.
Author contributions
====================
XG and HS designed and carried out experiments and analysis. XG wrote the manuscript. XG, QX and BL modified the manuscript. All authors have read and approved the final article.
The following online material is available for this article:
Table S1- Basic information of DNA methyltransferases from *Arabidopsis* and rice.Click here for additional data file.
Table S2- Primer pairs used in quantitative RT-PCR analysis.Click here for additional data file.
Table S3- Putative *cis-*elements enriched in the promoters of tomato MTases genes.Click here for additional data file.
Figure S1- Microarray expression data.Click here for additional data file.
| {
"pile_set_name": "PubMed Central"
} |
{.69}
{.70}
{.71}
{.72}
{.73}
{.74}
{.75}
| {
"pile_set_name": "PubMed Central"
} |
Genomics and particularly mutational profiling have deepened our understanding of cancer pathology and laid the foundation for precision therapies as a major field in modern oncology. Comprehensive genetic profiling has even led to proposals to replace the current histology-based WHO tumor typing with molecular classes based on the rationale that similar molecular alterations have similar clinical relevance across cancers \[[@R1]-[@R3]\]. However, targeted drugs against oncogenic mutations shown to be efficacious in one cancer are often ineffective in another \[[@R4]\]. But not only is it often not possible to transfer druggability of a mutation from one histological tumor type to another, response to targeted therapies shows significant variations also within the same tumor type. While some patients show long-term benefit, others quickly relapse or show no response to therapy despite identical actionable mutations. The reasons for this variability in biological behavior and therefore clinical relevance lie in the often complex genetic background in many tumors, but are also related to epigenetic, post-transcriptional and post-translational modifications not captured by mutational profiling alone. Non-small cell lung cancer, for instance, harbors over 800 genetic aberrations \[[@R5]\], on average, and although many of those are believed to be passenger mutations, a substantial number of mutations can be expected to be involved in the modulation of resistance mechanisms. However, the influence of these rare "tail" mutations is difficult to evaluate for two related reasons. First, even if a large number of mutations can be excluded based on prior biological knowledge, identifying drug combinations to overcome resistance is challenging even for a small number of mutations due to the combinatorial complexity (as an example, 780 alternative two-drug combinations exist for 40 targets \[[@R6]\]). Moreover, even if candidate targeted combination therapies are identified, patient recruitment for clinical trials is difficult due to the rarity of the druggable mutations. Therefore, ways beyond genomics have to be found to identify functionally and clinically relevant molecular alterations within the complex mutational landscape of cancer.
{#F1}
One way to assess the functional impact of genetic alterations in cancer is to study their relationship with corresponding (phospho-)proteomic profiles. Using publicly available proteogenomic profiling data across major cancer types, we propose a computational approach that systematically evaluates if and to what extent genetic aberrations observed in tumors are associated with distinct proteomic profiles \[[@R7]\]. Our analysis shows that distinct proteomic profiles are observed for mutated vs. wildtype genes in cancers known to be druggable for the respective mutations, which is not the case for mutations for which targeted therapies are ineffective. This approach therefore facilitates predicting potentially oncogenic and/or actionable genes in the context of different histological tumor types. While this shows the capacity of proteomics to complement genomic profiling with respect to assessing functional impact for gene mutations with sufficient frequencies to allow for computational evaluation, the ultimate goal of such approaches must be the prediction on the level of individual patients to support precision oncology.
To achieve this, we further propose a combined experimental and computational systems proteogenomics approach that allows for a reduction of the mutational complexity and facilitates the identification of functionally relevant molecular aberrations that can be exploited to overcome resistance against targeted (mono-)therapy in individual patients \[[@R8]\]. It combines mutational profiling with proteomics and uses experimental perturbation with targeted drugs against which resistance has developed. Cell or tissue culture models are treated with the respective inhibitors subsequent to which time-course discovery phosphoproteomics is performed in comparison to a standard. Because static proteomics data is similarly complex as the mutational profiles, they do not per se facilitate a better understanding of the pathological mechanisms, but dynamic changes in protein phosphorylation and expression analyzed after perturbation and compared with the t=0 status may offer a more specific picture. This is supported by bioinformatic methods including differential correlation \[[@R9]\], which helps identify relevant groups of phosphoproteins, as well as network models used to topologically relate mutational and proteomic profiles. Through these data analysis and integration steps, the complexity of the molecular profiling data can be substantially reduced to a short list of likely functionally relevant target molecules that can then be further evaluated experimentally and clinically.
In summary, the ability to interpret and assess the functional and clinical relevance of the increasingly comprehensive mutational profiling data accruing not only in research but also in clinical cancer diagnostics is reaching its limits when relying only on pre-clinical experimental data and knowledge on biological pathways. Moreover, the combinatorial complexity of potential druggable target combinations is incompatible with testing in classical clinical trials. As a solution, combining genomic with proteomic profiling and classical computational analysis as well as advanced machine learning techniques can contribute to identifying functionally and clinically relevant molecular alterations which in the future may complement molecular diagnostics for precision oncology.
| {
"pile_set_name": "PubMed Central"
} |
Dear Editor,
We present a case of glomus tumor located on the lower third of the right thigh in a patient who presented with pain in the knee region for many years. Investigation for osteomuscular disorders was not successful.
A 50-year-old man sought dermatological consultation reporting an intense, episodic shooting pain in the right knee for 4.5 years, that worsened with movement or mild touch, including with simple touch of clothes. He denied worsening with cold. He had seen many orthopedists with such complaint and underwent multiple imaging studies, among them 5 nuclear magnetic resonance (NMR) and 3 ultrasounds (US) of the knee and lower third of the right thigh. In only one US, a small subcutaneous nodule on the left thigh was described, however, such finding was not taken into consideration by the physicians then. The tumor was not described in any of the reports from NMR. In view of the undefined diagnosis, he still underwent electroneuromyography, which did not show any abnormalities.
The osteomuscular examination did not reveal edema, heat or crackling of the knee. Dermatological examination revealed a violaceous papule, measuring 4 mm, close to the femur lateral condyle, which triggered a sharp pain upon minimal contact ([Figure 1](#f1){ref-type="fig"}). Non-contact dermoscopic examination with polarized light showed a violaceous, homogeneous lesion with no structures or patterns ([Figure 2](#f2){ref-type="fig"}). Because of the clinical aspect and peculiar symptoms, we raised the hypothesis of extradigital glomus tumor and proceed to surgical excision.
Figure 1Violaceous papule on the left thigh Figure 2Non-contact polarized dermoscopy showing a violaceous, homogeneous, structureless and patternless lesion
Histopathology revealed a proliferation of blood vessels surrounded by small, uniform cells with round or oval-shaped nuclei (glomus cells), typical findings of glomus tumor ([Figure 3](#f3){ref-type="fig"}). Resection margins were free. The patient came for follow-up after 2 months and was completely asymptomatic.
Figure 3**A** - Encapsulated neoplasia in the subcutaneous tissue (Hematoxylin & eosin, X20). **B** - Benign neoplasia formed by cuboidal, uniform cells around vessels (Hematoxylin & eosin, X200)
Glomus tumor is considered a rare, benign neoplasia, that corresponds to approximately 1.6% of the soft tissue tumors located on the extremities.^[@r1]^ it arises from the glomus bodies, neuromyoarterial structures in the skin that function as specialized arteriovenous anastomoses, aiding in the regulation of body temperature and blood flow. These structures are found all over the skin, but are more concentrated in the extremities, what explains the higher involvement of the subungual region by the tumor.^[@r2]^ Glomus tumor is equally prevalent in both genders, however, the "ectopic" locations are more common in men.^[@r1]^ Clinically, it presents as a firm, purple-blue, usually single nodule, with the classic triad of pain, hyperesthesia on palpation and cold sensitivity (worsening of the pain when exposed to the cold).^[@r2],[@r3]^ When its location is extradigital, it frequently becomes a diagnostic challenge, because many times it does not present with the classical clinical picture. Schiefer *et al.*, in a study on extradigital glomus tumors, found 86% of localized pain and heat and only 1% of cold insensitivity, such as in the presented case (the patient denied worsening with the cold).^[@r3]^ Our patient had a lesion on an unusual site, close to the knee, what led him to different orthopedists, who also interpreted his complaints as osteomuscular in nature and recommended treatment with physiotherapy. This professional was the one who referred to a dermatology consultation.
There are few reports of glomus tumors mimicking joint disorders. Especifically on the knee, the tumor was already described in the subcutaneous tissue, in the patellar ligament and even in an intra-articular location. Since the differential diagnosis for knee pain is broad, when this neoplasia is present the diagnosis is rarely made, and it can take up to 40 years after the onset of symptoms.^[@r4]^
The diagnosis is suggested by the clinical presentation, since dermoscopy is non-specific. The findings described include homogeneous white structure with peripheral telangiectasias or structureless violaceous areas, as in our case.^[@r5]^ Imaging tests can also help in the diagnosis, being the NRM the most sensitive (82% to 90%), detecting lesions with a diameter smaller than 2 mm.^[@r2],[@r3]^ We assume that the lesion was not detected in our patient with the NRM because the complaint was of pain in the joint, so the radiologists did not recognize it either. Histology confirms the diagnosis, showing a tumor enveloped by a fibrous capsule, with vascular spaces surrounded by glomus cells (eosinophilic cytoplasm and oval or cuboidal nucleus).^[@r1]^
After surgical excision, there is complete regression of symptoms and low recurrence rates.^[@r2]^
We alert to the fact that painful skin and subcutaneous tissue tumors can mimic osteomuscular conditions, especially when the location is juxta-articular.
Financial support: None.
Work conducted at Hospital do Servidor Público Estadual de São Paulo, São Paulo (SP), Brazil.
Conflict of interest: None.
| {
"pile_set_name": "PubMed Central"
} |
Altered metabolism is a key characteristic of cancer cells. The metabolic flux in cancer cells is markedly reprogrammed to provide elevated amounts of building blocks for anabolic biosynthesis of macromolecules during rapid cell growth and proliferation[@b1][@b2]. In addition, cancer cells often possess enhanced capability to regulate redox homoeostasis to protect against oxidative stress in the tumor microenvironment[@b3][@b4]. However, the molecular mechanisms by which cancer cells sense metabolic signal to couple anabolic synthesis and redox homoeostasis to promote cancer cell proliferation and cell survival are not well understood.
Glucose flux through the glycolytic pathway can be diverted into the pentose phosphate pathway (PPP). The PPP plays a vital role in meeting the cellular demands for anabolic biosynthesis and providing anti-oxidative defense[@b5]. It generates ribose-6-phosphate for *de novo* synthesis of DNA and RNA, and the reducing equivalent reduced NADPH for reductive biosynthesis of lipids. NAPDH also functions as an important antioxidant for detoxification of high levels of reactive oxygen species (ROS) produced during rapid cell proliferation to promote cell survival. Activity of the PPP is known to be upregulated in cancer cells compared with normal epithelial cells[@b6]. Knocking down of key enzymes in the PPP inhibits tumor growth and sensitizes cancer cells to oxidative stress[@b7][@b8]. Glucose-6-phosphate dehydrogenase (G6PD) catalyses the first committed and rate-limiting step of the PPP. It catalyses the oxidation of G6P to 6-phosphogluconate and produces NADPH in the presence of NADP^+^. G6PD is considered the pacesetter of the PPP and the primary control point for NADPH production. G6PD activity is subjected to various regulatory mechanisms ranging from transcription to translation, further illustrating its importance in regulating cellular metabolism[@b5].
O-linked β-N-acetylglucosamine (O-GlcNAc) is a dynamic and inducible post-translational modification of serine and/or threonine residues of nuclear and cytosolic proteins[@b9]. In cells, a single set of antagonistic enzymes---O-GlcNAc transferase (OGT) and O-GlcNAc hydrolase are responsible for the addition and removal of GlcNAc moiety, respectively. O-GlcNAcylation has been identified in numerous proteins and shows a complex crosstalk with protein phosphorylation[@b10]. Increasing evidence has shown that O-GlcNAcylation serves important roles in regulating gene transcription, cellular signalling and stress responses[@b11]. Generally considered as a 'nutrient sensor\' of cells, recent findings also indicate that O-GlcNAcylation may actively and directly participate in regulating cellular metabolism[@b12]. O-GlcNAc levels are significantly elevated in various cancers. Aberrant O-GlcNAcylation has been shown to correlate with cancer cell proliferation, invasion and metastasis both *in vitro* and *in vivo*[@b13][@b14][@b15]. Thus, deregulation of O-GlcNAcylation appears to be a general feature of cancer cells. However, the detailed mechanisms by which protein-specific O-GlcNAcylation contributes to cancer metabolic reprogramming and tumorigenesis remain largely unknown.
Here, we present evidence that O-GlcNAcylation of G6PD coordinates cancer cell anabolic biosynthesis and redox homoeostasis to promote tumor growth *in vivo*. Mechanistically, glycosylation activates G6PD activity and increases glucose flux through the PPP, thereby generating precursors for nucleotide and lipid biosynthesis, and reducing equivalents for antioxidant defense. In addition, glycosylation of G6PD promotes cancer cell proliferation *in vitro* and tumor growth *in vivo*. Importantly, G6PD glycosylation is elevated in human lung cancers. Thus, our study identifies a previously unknown mechanism for the regulation of cancer metabolism and tumor growth by protein O-glycosylation.
Results
=======
G6PD is dynamically modified by O-GlcNAc at serine 84
-----------------------------------------------------
Previous proteomic studies[@b16][@b17] have revealed that a large number of metabolic enzymes including G6PD are modified by O-GlcNAc in eukaryotic cells, suggesting a critical role of O-GlcNAcylation in regulating cell metabolism. To investigate whether O-GlcNAcylation directly regulates the PPP, we focus on the rate-limiting enzyme G6PD. To confirm that G6PD is O-GlcNAcylated, we employed a well-established chemo-enzymatic labelling approach[@b16]. We enzymatically labelled all O-GlcNAc-modified proteins from 293T cell lysates with an azido-N-acetylgalactosamine sugar. Labelled proteins were then biotinylated via Cu(I)-mediated \[3+2\] azide-alkyne cycloaddition (CuACC) chemistry and captured with streptavidin-agarose beads. Subsequent immunoblotting of the captured proteins with an antibody against G6PD showed strong O-GlcNAcylation of G6PD ([Fig. 1a,b](#f1){ref-type="fig"}). Calculation of the glycosylated protein versus total G6PD protein yielded an estimate of basal glycosylation level to be 8±2.1%. Glycosylation of G6PD was enhanced by 4.5-fold in cells by overexpression of OGT, the enzyme responsible for the addition of O-GlcNAc onto proteins ([Fig. 1c](#f1){ref-type="fig"}). To corroborate with this result, we immunoprecipitated endogenous G6PD from 293T cell lysates and immunoblotted with a pan-anti-O-GlcNAc antibody (CTD110.6), which yielded a distinct signal albeit much lower sensitivity compared to the chemo-enzymatic approach ([Supplementary Fig. 1](#S1){ref-type="supplementary-material"}). Thus, G6PD is O-GlcNAc-modified in cells.
To identify the site(s) of O-GlcNAcylation on G6PD, we transiently co-expressed Flag-tagged G6PD and OGT in 293T cells. After immunoprecipitation and in-gel trypsin digestion of G6PD, peptides were subjected to high-resolution mass spectrometry analysis (nanoLC-LTQ-CID/ETD-Orbitrap). We identified a single site of O-GlcNAcylation at Serine 84, a highly conserved residue in homologues proteins among mammals ([Fig. 1d](#f1){ref-type="fig"} and [Supplementary Fig. 2](#S1){ref-type="supplementary-material"}). Mutation of Ser84 to valine abolished the glycosylation signal, supporting that Ser84 is the only site of O-GlcNAcylation on G6PD ([Fig. 1d](#f1){ref-type="fig"}). Notably, no other forms of modification such as phosphorylation were detected at the same serine residue.
Accumulating research data have suggested that protein O-GlcNAcylation is induced in response to various forms of cell stress[@b18], and that elevation of O-GlcNAcylation plays an important role in modulating critical biological pathways to improve cell survival[@b18]. To investigate whether O-GlcNAcylation of G6PD is dynamically regulated in cells, we subjected cells to different stress conditions and probed the O-GlcNAcylation levels of G6PD. Notably, G6PD glycosylation was induced in a time-dependent manner under hypoxic conditions (3.1- to 4.7-fold induction, [Fig. 1e](#f1){ref-type="fig"}). We found that hypoxic treatment induced global O-GlcNAc levels in cells and endogenous OGT expression ([Supplementary Fig. 3A](#S1){ref-type="supplementary-material"}). Consistently, *in vitro* enzymatic assays demonstrated that OGT activity in cell lysates was increased under hypoxia ([Supplementary Fig. 3B](#S1){ref-type="supplementary-material"}). Hypoxia is also known to induce profound changes in glucose metabolism, including increasing glucose uptake via the transcriptional upregulation of glucose transporters[@b19]. Indeed, the hypoxic treatment significantly enhanced glucose uptake rate in our study ([Supplementary Fig. 3C](#S1){ref-type="supplementary-material"}), nicely correlated with the induction of O-GlcNAcylation level. In addition, inhibition of glucose uptake by a small-molecule inhibitor suppressed the induction of G6PD O-GlcNAcylation under hypoxia ([Supplementary Fig. 3D](#S1){ref-type="supplementary-material"}). Thus, hypoxia induces G6PD glycosylation likely by increasing OGT expression and cellular glucose concentration, which serves as a biosynthetic precursor for O-GlcNAc. Similarly, G6PD glycosylation was also stimulated when cells were treated with high glucose concentration (4.2- to 5.8-fold induction), or with serum (1.7-fold induction; [Fig. 1f,g](#f1){ref-type="fig"}). In agreement with previous reports[@b3], growth factor (serum) stimulation significantly induced cellular glucose uptake rate in our study ([Supplementary Fig. 3E](#S1){ref-type="supplementary-material"}). Taken together, these results demonstrate that G6PD O-GlcNAcylation is dynamically regulated in response to different cellular conditions, suggesting a signalling role of G6PD glycosylation in cells.
O-GlcNAcylation of G6PD activates enzyme activity
-------------------------------------------------
To understand the biological significance of G6PD O-GlcNAcylation, we first examined the effect of O-GlcNAcylation on G6PD enzyme activity. Enhancing O-GlcNAcylation in 293T cells by OGT overexpression or pharmacological inhibition of O-GlcNAc hydrolase with a specific inhibitor thiamet-G[@b20] significantly increased G6PD enzyme activity by two to fourfold ([Fig. 2a](#f2){ref-type="fig"}). Mutation of S84 to valine (S84V) retained a similar activity as compared to the wild-type (WT) G6PD. However, the S84V mutant showed negligible response in enzyme activity on OGT overexpression or thiamet-G treatment ([Fig. 2a](#f2){ref-type="fig"}). Similar results were obtained when cells were subjected to hypoxic treatments to induce G6PD glycosylation ([Supplementary Fig. 4](#S1){ref-type="supplementary-material"}). Thus, these results suggest that S84 is an important regulatory site of G6PD activity.
To further understand the effect of S84 glycosylation on G6PD activity, we examined the steady-state kinetics of G6PD with different glycosylation levels. Specifically, we compared the kinetics of Flag-tagged WT G6PD expressed in 293T cells in the presence or absence of OGT overexpression. Flag-tagged G6PD with high O-GlcNAcylation displayed substantially higher catalytic efficiency (2.9-fold increase in *k*~cat~/*K*~m~ for NADP+ and 1.4-fold increase for G6P) than Flag-tagged G6PD with low O-GlcNAcylation ([Supplementary Table 1](#S1){ref-type="supplementary-material"}). Notably, the *K*~m~ for NADP^+^ showed about twofold decrease in G6PD with high glycosylation levels, indicating a higher binding affinity of NADP^+^ to the enzyme. Consistently, the calculated dissociation constant (*k*~d~) for NADP^+^ using the fluorescence titration assay was 2.6 μM for G6PD with high glycosylation levels, which is about threefold lower than G6PD with low glycosylation levels (7.3 μM). Thus, the kinetics study suggests that O-GlcNAcylation induces G6PD activity likely by enhancing the binding affinity of NADP^+^ to G6PD.
The G6PD protein exists in different oligomeric states, ranging from monomer, dimer, tetramer to hexamer. Studies have shown that only dimeric and tetrameric forms of G6PD are catalytically active[@b21][@b22]. To examine whether S84 glycosylation affects oligomerization states of G6PD, we performed a protein crosslinking experiment using glutaraldehyde. As shown in [Fig. 2b](#f2){ref-type="fig"}, enhancing O-GlcNAcylation levels by OGT overexpression resulted in substantial increased formation of dimeric G6PD, but not S84V mutant. Control immunoblots were shown in [Fig. 2c](#f2){ref-type="fig"}. These results suggest that O-GlcNAcylation activates G6PD in part by perturbing the equilibrium between different oligomeric forms to favour a higher oligomeric state.
O-GlcNAcylation of G6PD promotes the PPP
----------------------------------------
To create a system to study the effect of G6PD S84 glycosylation on cellular metabolism, we depleted endogenous G6PD and stably expressed small hairpin RNA (shRNA)-resistant Flag-tagged WT or S84V G6PD in A549 lung cancer cells (henceforth referred to as WT G6PD or S84V G6PD replacement cells, [Supplementary Fig. 5](#S1){ref-type="supplementary-material"}). As the first and rate-limiting step of the PPP, induction of G6PD activity by O-GlcNAcylation is expected to have an impact on the metabolic flux of the PPP. Indeed, we observed a significant increase in flux (∼2.1-fold) through the oxidative PPP, as measured by the amount of released ^14^CO~2~ from \[1-^14^C\]-glucose, when OGT was overexpressed in WT G6PD replacement cells. In contrast, in S84V G6PD replacement cells, PPP flux exhibited a modest increase on OGT overexpression ([Fig. 3a](#f3){ref-type="fig"}). The small increase in PPP flux is likely due to the inhibition of phosphofructokinase 1 (PFK1), a key regulatory enzyme in glycolysis pathway, by O-GlcNAcylation as shown in the previous study[@b23]. Nevertheless, this result demonstrates that S84 O-GlcNAcylation of G6PD induced glucose flux through the PPP. This observation was independently verified by a metabolic tracing experiment, in which cells were metabolically labelled with 1,2-^13^C-glucose and the relative isotopic enrichment accumulation of singly versus doubly ^13^C-labelled lactate was measured by liquid chromatography--mass spectrometry (LC--MS) analysis ([Fig. 3b](#f3){ref-type="fig"}). In addition, we showed that hypoxia induced similar upregulation in PPP flux and that the upregulation was dependent on G6PD glycosylation ([Supplementary Fig. 6](#S1){ref-type="supplementary-material"}).
Our previous investigation on O-GlcNAcylation of PFK1 has demonstrated that O-GlcNAcylation inhibits PFK1 enzyme activity, resulting in rerouting a portion of glucose flux through the PPP[@b23]. We then carried out experiments to further understand the differential contribution to PPP regulation by glycosylation of PFK1 and G6PD. PFK1 has three isoforms (PFKL, PFKP and PFKM) that show different regulatory response by O-GlcNAcylation with PFKM being the least responsive[@b23]. We generated A549 stable cell lines that harbour only single isoforms. In addition, these PFK1 isoforms were mutated to become glycosylation-deficient (referred to as PFKL-SA, PFKP-SA and PFKM-SA). Consistent with the previous study, PPP flux increased significantly in cells expressing WT PFK1 isoforms on OGT overexpression ([Supplementary Fig. 7](#S1){ref-type="supplementary-material"}). The expression of glycosylation-deficient mutants of PFK1 (PFKL-SA or PFKP-SA) resulted in a large suppression of PPP flux, but the expression of PFKM-SA mutant only resulted in a modest decrease in PPP flux ([Supplementary Fig. 7](#S1){ref-type="supplementary-material"}). Replacement of WT G6PD with S84V G6PD further suppressed PPP flux in all three cases. These results suggest that O-GlcNAcylation regulates the PPP activity through the coordinated action of PFK1 and G6PD. In cases where PFKL or PFKP expression is predominant, PFK1 glycosylation exerts more control in regulating the PPP flux. On the other hand, when PFKM expression is predominant, G6PD glycosylation plays a more dominant role in PPP regulation.
To further examine the impact of G6PD glycosylation on major metabolic pathways, we subjected cell extracts to LC--MS analysis to determine the relative enrichment of specific metabolites. On OGT overexpression, the metabolic profile of WT G6PD replacement cells demonstrated a general increase in PPP metabolites, as compared with S84V G6PD replacement cells ([Fig. 3c](#f3){ref-type="fig"}). Metabolites involved in glycolytic pathway and tricarboxylic acid cycle showed no significant difference between WT and S84V G6PD replacement cells ([Fig. 3d,e](#f3){ref-type="fig"}). Control experiments were performed in parental A549 cells on G6PD overexpression ([Supplementary Table 2](#S1){ref-type="supplementary-material"}). The metabolite patterns suggested that S84 glycosylation imparted a specific metabolic effect on glucose metabolism through the PPP.
G6PD glycosylation promotes cell proliferation and survival
-----------------------------------------------------------
Rapidly dividing cancer cells require three basic outputs from metabolism: energy in the form of ATP, building blocks for macromolecular synthesis and cellular assembly and molecules to maintain the proper cellular redox environment[@b24]. Upregulation of the PPP flux induced by O-GlcNAcylation of G6PD would provide cells with pentose sugars for nucleotides and nucleic acid biosynthesis, as well as reducing equivalents from NADPH for lipid biosynthesis and to combat oxidative stress. We compared DNA synthesis in WT G6PD or S84V G6PD replacement A549 cells under hypoxic conditions using the 5-bromo-2\'-deoxyuridine incorporation assay. DNA synthesis was significantly increased in WT G6PD replacement A549 cells compared with the S84V mutant replacement cells ([Fig. 4a](#f4){ref-type="fig"}). As a control, depletion of G6PD with shRNA led to a sharp inhibition of DNA synthesis. These results are consistent with the previous observation that hypoxia activates G6PD glycosylation, which leads to enhanced flux through the PPP for nucleic acid biosynthesis. Control experiments were carried out under normoxia ([Supplementary Fig. 8](#S1){ref-type="supplementary-material"}). In addition, we found that S84V mutant replacement cells showed decreased lipogenesis compared with WT replacement cells under hypoxic conditions ([Fig. 4b](#f4){ref-type="fig"}). Consistently, WT G6PD replacement cells exhibited significantly higher cell proliferation rate than S84V mutant replacement cells under both hypoxic and normoxic conditions, yet with more pronounced effect under hypoxia ([Fig. 4c](#f4){ref-type="fig"}, [Supplementary Fig. 9](#S1){ref-type="supplementary-material"}). As a control, depletion of G6PD in A549 cells inhibited cell proliferation. Addition of Nuc (four ribonucleotides and four deoxyribonucleotides) to the culture medium partially rescued cell proliferation defect in S84V G6PD replacement cells ([Fig. 4c](#f4){ref-type="fig"}, [Supplementary Fig. 9](#S1){ref-type="supplementary-material"}). Thus, these results suggest that O-GlcNAcylation of G6PD enhances cellular biosynthesis and promotes cell proliferation.
Consistent with increased glucose flux through the PPP, enhancing O-GlcNAc levels by OGT overexpression in WT G6PD replacement cells resulted in 1.8- and 2.6-fold increase in NADPH and reduced glutathione (GSH) levels, respectively ([Fig. 4d,e](#f4){ref-type="fig"}). Blocking glycosylation of G6PD significantly suppressed the induction of NADPH and GSH, demonstrating the importance of O-GlcNAc glycosylation at S84 in regulating NADPH homoeostasis. To further confirm that glycosylation of G6PD plays an important role in antioxidant defense, we measured the sensitivity of A549 replacement cells to ROS production and ROS-mediated cell death. Induction of ROS levels by diamide was significantly suppressed in WT G6PD replacement cells as compared with S84V mutant replacement cells ([Fig. 4f](#f4){ref-type="fig"}). ROS levels were significantly induced by hypoxic treatments in a larger extent in S84V mutant replacement cells ([Supplementary Fig. 10](#S1){ref-type="supplementary-material"}). Consistently, compared with WT replacement cells, S84V mutant replacement cells were markedly more sensitive to hydrogen peroxide (H~2~O~2~) treatment, and exhibited higher percentage of ROS-induced cell death ([Fig. 4g](#f4){ref-type="fig"}). Addition of GSH, the ROS scavenger, to the culture medium partially rescued cell proliferation defect in S84V G6PD replacement cells ([Fig. 4c](#f4){ref-type="fig"}, [Supplementary Fig. 9](#S1){ref-type="supplementary-material"}). Combination of GSH and Nuc appeared to augment the rescue effect. These results indicate that G6PD O-GlcNAcylation plays an important role in regulating redox homoeostasis to promote cancer cell survival and cell proliferation.
G6PD glycosylation promotes tumor formation
-------------------------------------------
To gain a better understanding of the impact of G6PD glycosylation on cancer development, we first examined whether G6PD is glycosylated in different human solid tumor cells, including breast cancer cell line, MCF7, lung cancer cell lines, H661 and A549, ovarian cancer cell line, SKOV-3, melanoma cell line, A375, and osteosarcoma cell line, U2OS ([Supplementary Fig. 11](#S1){ref-type="supplementary-material"}). G6PD glycosylation was observed in all of these cancer cell lines even though the glycosylation levels were varied, suggesting that G6PD glycosylation is potentially linked to tumor pathology.
To determine whether G6PD glycosylation is important for tumor formation *in vivo*, we injected WT G6PD or S84V G6PD replacement A549 cells into immunocompromised mice and assayed their ability to form tumours. Mice injected with S84V G6PD replacement cells showed a significant delay in tumor formation compared with mice injected with WT G6PD replacement cells, producing tumours with much smaller total mass ([Fig. 5a](#f5){ref-type="fig"}). Western blotting analysis confirmed that Flag-tagged WT or S84V G6PD proteins were retained in tumours and that WT G6PD was O-GlcNAcylated *in vivo* ([Supplementary Fig. 12](#S1){ref-type="supplementary-material"}). These results are consistent with the *in vitro* cell proliferation data and demonstrate that glycosylation of G6PD at S84 provides a critical growth advantage to tumours *in vivo*.
G6PD glycosylation is upregulated in lung cancers
-------------------------------------------------
Cellular O-GlcNAcylation is reported to be upregulated in various human cancers, indicating that targeting O-GlcNAcylation may serve as a novel anti-cancer strategy[@b25]. The finding that G6PD O-GlcNAcylation is important for A549 lung cancer cell proliferation and tumor growth prompted us to examine G6PD glycosylation in human lung cancers. We obtained a total of 39 pairs of primary human lung cancer tissue samples with matched adjacent normal lung tissues, and determined the level of glycosylation by the chemoenzymatic tagging approach and normalization to G6PD protein levels. Among these samples, eight pairs showed minimal G6PD expression in either cancer or normal tissues that prevented reliable quantification. Among the remaining 31 pairs of samples, 16 pairs showed relatively higher level of glycosylated G6PD in cancer tissues than the matched normal tissues ([Fig. 5b](#f5){ref-type="fig"} and [Supplementary Fig. 13](#S1){ref-type="supplementary-material"}). Quantification of these samples confirmed that the increase in the ratio of glycosylated versus total G6PD proteins is statistically significant ([Fig. 5c](#f5){ref-type="fig"}). Previous studies showed that O-GlcNAcylation and OGT expression were significantly elevated in different cancer tissues compared with the adjacent normal tissues, including lung cancer[@b13][@b14]. Thus, we further analysed the expression levels of OGT protein in these samples. Among 31 pairs of samples, 20 pairs showed relatively higher level of OGT expression in cancer tissues than in the matched normal tissues ([Fig. 5b,d](#f5){ref-type="fig"}, [Supplementary Fig. 14](#S1){ref-type="supplementary-material"}). The samples that showed higher level of G6PD glycosylation also showed higher level of OGT expression. Thus, the increased G6PD glycosylation correlates with the increased OGT expression in lung cancer.
The development of non-small cell lung cancer is often divided into four stages (I, II, III and IV) based on tumor size and metastatic features. To determine whether G6PD glycosylation is associated with lung cancer progression, we analysed the level of glycosylation in the previous 16 paired samples according to their stages. The levels of G6PD glycosylation were increased significantly in all stages when compared with the normal tissues ([Fig. 5e](#f5){ref-type="fig"}). However, there was no significant difference in the levels of glycosylation among different stages ([Fig. 5e](#f5){ref-type="fig"}). Collectively, these results suggest that G6PD glycosylation may play a role in lung cancer initiation, but not progression of the disease.
Discussion
==========
Increasing evidence points to pivotal roles for O-GlcNAcylation in regulating a diverse set of proteins that function in transcription, insulin signalling and cellular stress response[@b26]. Our studies identify a previously unknown mechanism by which this modification regulates metabolic flux to promote proliferation and survival of cancer cells. Activation of G6PD activity via O-GlcNAcylation upregulates glucose flux through the PPP, leading to increased production of precursors involved in the biosynthesis of nucleotides. Enhancing flux through the PPP also provides reducing power in the form of NADPH and GSH to combat ROS, thereby supporting cell survival under oxidative stress. The dynamic induction of G6PD glycosylation in response to hypoxia and nutrient levels and coordination of metabolic reprogramming for cancer cell proliferation suggest that O-GlcNAcylation not only serves as a cellular sensor of metabolic states, but also constitutes a key metabolic regulator of glucose flux. Our study also provides a new mechanistic insight into the expanding scope of redox homoeostasis in cells.
It has long been established that the activities of metabolic enzymes are regulated by small-molecule metabolites. In contrast to the non-covalent, often transient, interactions of allosteric modulators, the O-GlcNAcylation is covalent, yet reversible and dynamically responsive to metabolic status. In addition, unlike the small molecules UDP-GlcNAc and O-GlcNAc are not primary or secondary metabolites. This may represent a difference in the function of O-GlcNAc compared with those small molecules. Given that many metabolic enzymes have been recently shown to be acetylated[@b27][@b28][@b29], phosphorylated[@b30][@b31] and glycosylated[@b17][@b23], the post-translational modification of metabolic enzymes may serve as a general mechanism for controlling cellular metabolism.
Elevated PPP flux has been demonstrated in cancer cells[@b5]. In addition to providing precursors for anabolic biosynthesis and maintaining cellular redox homoeostasis, elevated PPP activity is also linked to cancer cell invasion, metastasis and resistance to cancer therapies. A plethora of mechanisms including activation of oncogenic signalling pathways and inactivation of tumor suppressors have been shown to regulate the PPP flux in a tightly controlled fashion[@b6]. More recent studies demonstrate that protein lysine acetylation positively or negatively regulates activities of key enzymes in the PPP, providing a selective metabolic requirement for cell proliferation and cell survival under oxidative stress[@b32][@b33]. Our findings that O-GlcNAcylation positively regulates G6PD activity add a new mechanistic insight into the regulation of PPP, and suggest that modulating G6PD activity may represent a potential therapeutic strategy for cancer.
Methods
=======
Cell culture and tumor tissues
------------------------------
The cell lines 293T, A549, MCF7, H661, SKOV-3, A375 and U2OS were obtained from ATCC and cultured according to ATCC protocols. Lung tumor tissues and matching tumor-adjacent normal tissues from the same patient were obtained from the Zhejiang Cancer Hospital Bio-specimen Repository (Hangzhou, China). Informed consent was obtained from the patients. Procedures related to human subjects were approved by the Ethic Committee of Zhejiang Cancer Hospital.
Immunoblotting
--------------
Cells were lysed in SDS lysis buffer (1% SDS, 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, and Complete protease inhibitors (Roche)), and the lysate was resolved on a 4--12% SDS-PAGE gel, transferred to nitrocellulose membrane, and immunoblotted with the indicated antibodies. Antibodies used in this study were obtained from the following sources: anti-O-GlcNAc antibody (clone 18B10.C7, Thermo Scientific, 1:1,000 final dilution), anti-G6PD antibody (clone G12, Santa Cruz Biotechnology, 1:1,000), anti-Flag antibody (clone M2, Sigma-Aldrich, 1:5,000 final dilution) anti-GAPDH antibody (clone G6, Santa Cruz Biotechnology, 1:1,000 final dilution). All protein concentrations were measured using the Bicinchoninic Acid protein assay (Pierce). Western blots were visualized and quantified using an Odyssey Infrared Imaging System (LI-COR Biosciences, Version 2.1). The uncropped blots for main figures were presented in [Supplementary Figs 15 and 16](#S1){ref-type="supplementary-material"}.
Generation of stable cell lines
-------------------------------
To generate the G6PD rescue A549 cell lines, Flag-tagged WT G6PD (cDNA clone obtained from Origene) and S84V G6PD were cloned into the expression vector pLenti-FlagN-shRNA[@b34]. This vector allows for expression of an exogenous gene with the simultaneous knockdown of the endogenous gene. To knockdown endogenous G6PD, the shRNA sequence 5′-CCGGGCTGATGAAGAGAGTGGGTTTCTCGAGAAACCCACTCTCTTCATC-AGCTTTTTG-3′ (obtained from the Public TRC Portal database) or the corresponding scramble sequence 5′-CCGGTCCTAAGGTTAAGTCGCCCTCGCTCGAGCGAGGG-CGACTTAACCTTAGGTTTTTG-3′ was inserted into the same vector. The G6PD sequence was made resistant to the G6PD shRNA by introducing silent mutations (lower case: 5′-GtTGATGAAGcGAGTGGGTTTC-3′) using the enzyme PrimeSTAR Max DNA Polymerase (Takara). Lentiviruses were produced from these constructs using a three-plasmid packaging system as described[@b23]. A549 cells were infected with the lentiviruses and selected for monoclonal cells with green fluorescence.
Analysis of G6PD glycosylation
------------------------------
Chemoenzymatic labelling and biotinylation of proteins in cell lysates were carried out as described previously[@b16] Briefly, cell lysate (500 μg) was labelled according to the Click-iT O-GlcNAc Enzymatic Labelling System protocol (Invitrogen), and conjugated with an alkyne-biotin compound as per the Click-iT Protein Analysis Detection Kit protocol (Invitrogen). Control experiments were carried out in parallel in the absence of the labelling enzyme GalT or UDP-GalNAz. Biotinylated lysates were precipitated using methanol and chloroform as described in the Click-iT Protein Analysis Detection Kit protocol, resolubilized in 1% SDS, and neutralized with an equal volume of neutralization buffer (6% NP-40, 100 mM Na~2~HPO~4~, 150 mM NaCl). Lysates were then incubated with strepavidin resin (Pierce) with end-to-end rotation at 4 °C overnight. Resin was then washed five times with 1 ml of low-salt buffer (100 mM Na~2~HPO~4~, 150 mM NaCl, 0.1% SDS, 1% Triton X-100, 0.5% sodium deoxycholate) and five times with 1 ml of high-salt buffer (100 mM Na~2~HPO~4~, 500 mM NaCl, 0.2% Triton X-100). Biotinylated proteins were eluted by boiling the resin in 50 mM Tris-HCl pH 6.8, 2.5% SDS, 100 mM DTT, 10% glycerol and 20 mM biotin for 10 min. Western blotting analysis was carried out with anti-G6PD or anti-Flag antibodies.
To quantify the level of glycosylation, the intensity of the total G6PD protein band (Input) and the glycosylated G6PD protein band (Elution) were measured, and the ratio of the intensity of the glycosylated protein versus the intensity of the total protein was taken as the level of glycosylation.
G6PD purification and enzymatic assays of G6PD and OGT
------------------------------------------------------
G6PD expression constructs were individually transiently transfected into 293T cells and allowed for expression for 48 h. To purify Flag-tagged G6PD, cells were lysed in Triton X-100 lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 5 mM thimet-G and Complete protease inhibitor cocktail). The lysate (7 mg) was diluted to 2 mg ml^−1^ with NETFS buffer (100 mM NaCl, 50 mM Tris-HCl pH 7.4, 5 mM EDTA, 5 mM thiamet-G and Complete protease inhibitor cocktail). The sample was incubated with anti-Flag M2 affinity gel (400 μl; Sigma-Aldrich) at 4 °C overnight with end-to-end rotation. The gel was then washed twice with 10 ml of NETFS containing 1% Triton X-100, and twice with 10 ml of NETFS. The Flag-G6PD protein was eluted with the 3 × Flag peptide (Sigma-Aldrich) in NETFS buffer according to the manufacturer\'s protocol. The eluent was further purified and concentrated using an Amicon Ultra Centrifugal Filter (10 kDa molecular weight cutoff; Millipore) in a buffer containing 50 mM Tris-HCl pH 7.5, 100 mM KCl, 5 mM MgCl~2~ and 5% glycerol.
G6PD activity was measured using purified Flag-tagged G6PD in the reaction buffer containing 55 mM Tris-HCl pH 7.8, 3.3 mM MgCl~2~, 6 mM NADP, and 0.1 M G6P. The reaction mixture was incubated at 30 °C for 2 minutes before initiated by adding the G6PD enzyme (0.01 mg ml^−1^). Absorbance was recorded at 340 nm every 15 s for 10 min using a UV--vis spectrophotometer (SHIMADZU UV-2550). One unit of G6PD activity (IU) is defined as the amount of enzyme that catalyses the conversion of 1 μmol of NADP to NADPH per minute at 30 °C. Since absorbance ΔA (340 nm) of 1 μmol ml^−1^ of NADPH is 6.22 in a 1-cm light path, G6PD enzyme activity is calculated as:
Assays to determine steady-state kinetic parameters were performed in the above-mentioned buffer with varying amounts of G6P and NADP. To ensure initial rate measurements, the amount of G6PD was adjusted to give a linear increase in readings within the first 2 min of the reaction. Kinetic parameters were calculated using the equation as described[@b35].
OGT activity was determined using a modified coupled enzyme assay as described[@b36] Cell lysates were concentrated to 10 mg ml^−1^ in centricon and used as OGT enzyme source. Arbitrary absorbance was normalized to total protein concentrations.
Hypoxic and glucose treatment
-----------------------------
Hypoxia experiments were performed in a sealed hypoxia chamber (Proox Model 110, BioSpherix, Ltd.) filled with 1% O~2~, 5% CO~2~, and 94% N~2~ at 37 °C and 60% cell confluency for the indicated periods of time. For glucose treatment experiments, A549 stable rescue cells were seeded at a density of 1 × 10^5^ cells per ml in a 10-well tissue culture plate. The cells were cultured in low glucose DMEM media (5 mM glucose) for two passages before switched to high glucose DMEM media (25 mM glucose) for the indicated periods of time.
Glucose uptake assay
--------------------
Glucose uptake assay was performed using the Glucose Uptake Fluorometric Assay Kit (BioVison) according to the manufacturer\'s instructions. Briefly, cells were seeded at 2,500 cells per well at a 96-well tissue culture plate. Cells were washed with phosphate-buffered saline (PBS) and incubated in serum-free DMEM medium for 2 h, and further incubated in hypoxic chambers for indicated periods of time. Cells were then quickly washed with PBS and glucose uptake was initiated by incubating in KRPH buffer (20 mM Hepes, 5 mM KH~2~ PO~4~, 1 mM MgSO~4~, 1 mM CaCl~2~, 136 mM NaCl, 4.7 mM KCl, pH 7.4) containing 10 mM 2-deoxyglucose for 20 min. Cells were then washed with PBS three times and proceeded to oxidation step to generate recordable fluorescence. Arbitrary fluorescence counts were normalized to total protein concentrations. For inhibitor treatment, cells were pretreated with 100 μM phloretin for 30 min before incubated with 2-deoxyglucose.
Site mapping of G6PD glycosylation
----------------------------------
Flag-tagged G6PD and HA-tagged OGT were co-transfected in 293T cells. After 48 h, Flag-tagged G6PD was isolated using anti-Flag M2 agarose beads as described above. The bound protein was eluted in a buffer containing 4% SDS and 100 mM Tris-HCl, pH 8.0. After SDS-PAGE (4--12% Bis--Tris gels) and staining with Bio-Safe Coomassie blue R250 (0.25%) Stain (Bio-Rad). The G6PD protein band was excised and manually digested in-gel with trypsin. The extracted peptides were lyophilized and reconstituted in 1 × binding buffer (Glycoprotein Isolation Kit WGA, Thermo Scientific) and incubated with WGA lectin resin (Glycoprotein Isolation Kit WGA, Thermo Scientific) at 4 °C with end-to-end rotation overnight. The resin was washed according to the manufacturer\'s protocol, and the bound peptides were eluted with the provided elution buffer. The eluent was further purified by reverse-phase HPLC (Agilent 1100) using a preparative reverse-phase column (Agilent Eclipse XDB-C18; 5 μm, 9.4 × 250 mm) and a gradient of 5--30% B buffer over 20 min at 4 ml min^−1^ (A buffer, 0.5% aqueous AcOH; B buffer, 100% MeCN). Fractions eluting between 5--12 min were collected, pooled, lyophilized and subjected to nanoLC-LTQ-CID/ETD-MS analysis on an LTQ-Orbitrap Velos as previously described[@b37]. Data search were performed by Proteome Discovery (MASCOT search engine, version 1.3) with O-GlcNAc (Ser/Thr) set as variable modification.
Analysis of the PPP activity
----------------------------
PPP activity was determined by following the procedure as described[@b23]. Briefly, 2 × 10^6^ cells (WT or S84V G6PD A549 replacement cells, with or without OGT overexpression) were grown in a 6-cm culture plate in sodium bicarbonate-free RPMI medium supplemented with 10% FBS, 20 mM HEPES, 5 mM glucose and 0.2 μCi of \[1-^14^C\]-glucose or \[6-^14^C\]-glucose (American Radiolabeled Chemicals). The cells were placed in a closed glass vial, the centre of which was covered with filter paper soaked in 100 μl of 5% KOH, and incubated at 37 °C for 4 h. The filter paper was removed and the radioactivity was determined using a LS 6500 Multi-Purpose Scintillation Counter (Beckman Coulter). PPP activity was calculated as the difference between the radioactivity levels of samples obtained from \[1-^14^C\]-glucose and samples obtained from \[6-^14^C\]-glucose, normalized to cell number.
Determination of PPP flux by stable isotope labelling was followed as described[@b23]. Briefly, WT or S84V G6PD A549 replacement cells (with or without OGT overexpression) were seeded at a concentration of 200,000 cells per well in a 6-well tissue culture plate and grown in complete RPMI 1640 culture medium for 24 h. Cells were pulse labelled in RPMI 1640 supplemented with 5 mM \[1,2-^13^C\]-glucose and 2 mM glutamine for 4 h, and subjected to metabolite extraction with 80% aqueous methanol and dried by speedvac. Cell extracts were then analysed for relative abundance of ^13^C-metabolites by liquid chromatography-triple quadrupole mass spectrometry using scheduled selective reaction monitoring for each metabolite of interest, with the detector set to negative mode as described[@b23]. Extracted metabolite concentrations were calculated from standard metabolite build-up curves using natural ^12^C synthetic metabolites and normalized against cell number as well as the internal ^13^C-labelled metabolite standards added at the time of metabolite extraction. Calculations for relative percentage of PPP flux were as described[@b38].
Determination of NADPH and GSH levels
-------------------------------------
NADPH levels were determined using a colorimetric NADP^+^/NADPH Quantitation Kit (BioVision) according to the manufacturer\'s protocol. The signal at 450 nm was recorded using a Victor 3 microplate reader and normalized to protein concentration. GSH levels were measured using a Glutathione Assay Fluorimetric Kit (Sigma-Aldrich) according to the manufacturer\'s procedure. Fluorimetric signal was recorded on a Victor 3 microplate reader and normalized to protein concentration.
LC--MS for metabolite analysis
------------------------------
The profiling of representative metabolites in PPP, glycolysis and tricarboxylic acid cycle were carried out on an Xevo TQ-S tandem mass spectrometer equipped with an electrospray source operating in the negative-ion multiple-reaction monitoring mode. The ion source settings were as follows: ion spray voltage, −3500 V; source temperature, 120 °C; desolvation gas, cone gas and nebulizer gas at settings 600 l h^−1^, 150 l h^−1^ and 7 Bar, respectively. Chromatographic separation was performed on a Phenomenex Luna Amide column (5 μm, 100 × 2.0 mm). The quantitative multiple-reaction monitoring transition of each particular metabolite was described as below, *m/z* 259\>97 for G6P; *m/z* 259\>97 for fructose-6-phosphate (F6P); *m/z* 339\>97 for fructose-1,6-diphosphate; *m/z* 169\>97 for dihydroxy-acetone-phosphate (DHAP) and glyceraldehyde-3-phosphate; *m/z* 185\>79 for 2-phosphoglycerate and 3-phosphoglycerate (2PG/3PG); *m/z* 167\>79 for phosphoenolpyruvate; *m/z* 87\>43 for pyruvate; *m/z* 808\>79 for acetyl-coenzyme A; *m/z* 191\>111 for citrate; *m/z* 191\>73 for isocitrate; *m/z* 145\>101 for 2-oxoglutarate; *m/z* 117\>73 for succinate; *m/z* 115\>71 for fumarate; *m/z* 133\>115 for malate; *m/z* 131\>87 for oxaloacetate; *m/z* 173\>85 for *cis*-aconitate; *m/z* 275\>79 for 6-phosphogluconate; *m/z* 229\>97 for ribose-5-phosphate, ribulose-5-phosphate and xylulose-5-phosphate; *m/z* 289\>97 for sedoheptulose-7-phosphate; *m/z* 199\>97 for erythrose-4-phosphate. Other MS/MS parameters, such as declustering potential and collision cell energy, were optimized for each particular metabolite[@b39]. Data were acquired and processed using MassLynx software (Version V4.1). Peak areas of individual metabolites were normalized against the total protein amount. The fold changes of the relative level of targeted metabolites are calculated (Data are presented as the mean±s.d. of triplicate biological experiments). The mass spectrometry raw data have been deposited in the iProX database.
Measurement of lipid synthesis
------------------------------
Cellular lipid synthesis was measured by incorporation of ^14^C-glucose. Cells were seeded at subconfluency on a six-well tissue culture plate. Cells were labelled with 4 μCi ml^−1^ of ^14^C-glucose for 2 h, followed by washing twice with cold PBS. Lipids in cells were extracted with 800 μl of hexane:isopropanol (3:2, v/v). Extracts were dried by speedvac, resuspended in 50 μl of chloroform, and subjected to scintillation counting. Counts were normalized with cell numbers.
Measurement of ROS levels and H~2~O~2~-mediated cell death
----------------------------------------------------------
A redox-sensitive dye 5(6)-chloromethyl-2\'7\'-dichlorodihydrofluorescein diacetate-acetyl ester (CM-H~2~DCFDA; Molecular Probes) was used to measure ROS levels in cell lines as previously described[@b23]. Briefly, cells were cultured overnight at a density of 4 × 10^5^ cells per well in a 12-well tissue culture plate in complete DMEM media. After treating cells with various concentrations of diamide (Sigma-Aldrich) as indicated for 10 min, fresh culture media were added. The cells were then incubated with 10 μM CM-H~2~DCFDA for 20 min and rinsed three times with PBS. The cells were lysed in 1% SDS and sonicated. The mixture was centrifuged (15,000 × *g*, 2 min) to remove any cell pellets. 100-μl aliquots were taken, and the fluorescent intensity was measured on a Victor 3 microplate reader. Signal intensity was normalized to protein concentration.
The percentage of cell death was measured using a lactate dehydrogenase (LDH)-based toxicology assay kit (Sigma-Aldrich) according to the manufacturer\'s protocol. The percentage of cell death was determined by comparing the amount of cytoplasmic LDH released into the culture medium relative to the total cytoplasmic LDH, as determined by the reduction of NAD^+^.
Cell proliferation analysis
---------------------------
Cell proliferation assays were performed by seeding 2,000 cells per well in a 96-well tissue culture plate in normoxia. Twenty-four hours after seeding, cells that were used for hypoxic treatments were further cultured in the sealed hypoxic chamber as mentioned above. Cell proliferation was determined by counting cell numbers as stained by trypan blue every 24 h for 96 h.
Xenograft studies
-----------------
Immunocompromised mice (BALB/c-nude, male, 5--6 week old, Charles River Laboratories) were injected subcutaneously with 4.5 × 10^6^ cells each (resuspended in 100 μl of serum-free F12-K medium) in the right flank (WT G6PD A549 replacement cells) and the left flank (S84V G6PD A549 replacement cells). Tumor growth was monitored every 3 days over a 7-week period. At the end of the seventh week, the tumours were harvested and weighed. Experiments were performed in accordance with the Zhejiang University Institutional Animal Care and Use Committee guidelines.
Statistical analysis
--------------------
*P* values were calculated from Student\'s paired *t*-test when comparing within groups and from Student\'s unpaired *t*-test when comparing between groups. For those analyses where more than one *t*-test is applied to the same data set, statistical analysis was performed by one-way analysis of variance and Bonferroni comparison post-test.
Additional information
======================
**Accession code:** The mass spectrometry data have been deposited in the iProx database under accession code IPX00037500.
**How to cite this article:** Rao, X. *et al.* O-GlcNAcylation of G6PD promotes the pentose phosphate pathway and tumor growth. *Nat. Commun.* 6:8468 doi: 10.1038/ncomms9468 (2015).
Supplementary Material {#S1}
======================
###### Supplementary Information
Supplementary Figures 1-16 and Supplementary Tables 1-2
This work was supported by the National Science Foundation of China (NSFC, grant nos. 31270865, 31322019, 81302739 and 31570804), the Thousand-Young-Talents Recruitment Program, the Outstanding Young Scholar Program of Zhejiang Province (LR15C050001), and the Fundamental Research Fund for the Central Universities. We also acknowledged the support from the Development Fund for Collaborative Innovation Center of Glycoscience of Shandong University.
**Author contributions** W.Y. and X.D. designed and directed the project. X.R. performed the experiments except where otherwise noted. X.D. performed the metabolic LC--MS analysis and glycosylation site mapping. L.Z., X.L. and Q.L. generated stable cell lines and peformed glycosylation analysis in cancer cell lines. W.M., Z.Z., H.X., Y.W., P.G.W. and B.S. analysed data. W.Y. wrote the paper and all authors participated in editing it.
{#f1}
{#f2}
![O-GlcNAcylation of G6PD regulates metabolic flux through the PPP.\
(**a**) PPP activity in WT and S84V G6PD replacement cells in the absence or presence of OGT overexpression, as determined by the amount of released ^14^CO~2~ from \[1-^14^C\]-glucose (*n*=3 assays). (**b**) Percentage of glucose flux through the PPP in WT and S84V G6PD replacement cells in the absence or presence of OGT overexpression, as measured by the relative accumulation of singly ^13^C-labelled lactate from cells metabolically labelled with 1,2-^13^C-glucose using reverse-phase triple-quadrupole LC--MS (*n*=3 assays). (**c**--**e**) Targeted analysis of abundance of different metabolites in major glucose metabolic pathways: PPP (**c**), glycolysis (**d**), and TCA cycle (**e**) in WT and S84V G6PD replacement A549 cells on OGT overexpression (*n*=3 experiments). Error bars denote mean±s.d. Statistical analysis was performed by Student\'s *t*-test (\**P*\<0.05).](ncomms9468-f3){#f3}
{#f4}
{#f5}
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Enhanced recovery after surgery (ERAS) programs, also known as fast-track surgery, are multimodal perioperative programs that aim to accelerate recovery, shorten hospital stay, and reduce postoperative complications. Postoperative management includes epidural anesthesia, early mobilization, early enteral nutrition, and early removal of catheters \[[@CR1]--[@CR3]\]. Early mobilization has been widely practiced and is an important component in preventing complications, including orthostatic hypotension (OH) during postoperative care \[[@CR4], [@CR5]\]. However, postoperative patients are frequently exposed to prolonged immobilization. Immobility has an important role in the development of neuromuscular weakness, atelectasis, insulin resistance, joint contractures, and OH \[[@CR6]--[@CR8]\]. In cardiothoracic and abdominal surgery, postoperative complications remain major clinical problems, despite advances in surgical techniques and perioperative care.
Postoperative OH is characterized by symptoms of dizziness, nausea, vomiting, or syncope during sitting or standing \[[@CR9]\]. OH is a well-known clinical complication that can delay early mobilization, although relatively little data are available regarding its mechanism and possible treatment \[[@CR10]\]. In addition, the pathophysiology of OH might be related to impaired cardiovascular regulation postoperatively \[[@CR11]\], but this relationship is not clearly understood. Previous studies have documented a 12-19% rate of incidence of OH during early postoperative mobilization in patients who were treated for breast cancer, had undergone hip arthroplasty, or had received some type of gynecological treatment \[[@CR12]--[@CR15]\]. Only one study assessed the cardiovascular response and orthostatic intolerance to early mobilization after video-assisted thoracic surgery (VATS). This study demonstrated a 35% incidence of OH \[[@CR16]\].
More surgical stress is anticipated in cardiothoracic and abdominal surgery compared to those surgeries. Although, incidence of OH after cardiothoracic and abdominal surgery has not been well-known, we hypothesize that more OH should be identified after cardiothoracic and abdominal surgery. The primary aim of this study was to examine the incidence of OH during early mobilization after major surgery. In addition, we investigated the independent risk factors associated with OH.
Methods {#Sec2}
=======
Patients {#Sec3}
--------
As a prospective observational study, we enrolled patients who underwent cardiac surgery (cardiac group), thoracic surgery (thoracic group), and abdominal surgery (abdominal group) at Nagasaki University Hospital and Seirei Mikatahara General Hospital from October 2008 to January 2012. In each group covered in cardiac group (e.g., coronary heart disease, valve diseases), in thoracic group (e.g., lung cancer, mediastinal neoplasm), in abdominal group (e.g., gastric cancer, liver and pancreatic cancer), respectively. All patients received standard perioperative management and nursing care following the critical path. Patients were included if they were older than 18 years, were undergoing planned surgery, and could provide written informed consent. Patients were established to have a performance status of 0 and to be clinically stable before the surgery. Patients were excluded if they had comorbid conditions that affected exercise performance (e.g., musculoskeletal or neurological impairment); had undergone re-operation; off-pump cardiac surgery; were experiencing atrial fibrillation, vomiting, and diarrhea before mobilization; or had a history of any cerebrovascular disease and postoperative admission to the intensive care unit. The Human Ethics Review Committee of Nagasaki University Hospital (Approval number: 09022760 − 2) and Seirei Mikatahara General Hospital (Approval number: 09-02) approved this study.
Measurement {#Sec4}
-----------
A physiotherapist evaluated cardiovascular response and symptoms during an attempt to perform a standing exercise on postoperative day 1 (POD 1). Systolic blood pressure (mmHg; SBP), diastolic blood pressure (mmHg; DBP) and mean arterial pressure (mmHg; MAP), heart rate (beats per minute; HR), oxygen saturation (percentage; SpO~2~), and respiratory rate (frequency per minutes; RR) were monitored throughout mobilization using the BSM-2300 series Life scope-i (Nihon Kohden Corporation, Tokyo, Japan). Their values in a supine position were defined as the baseline levels. We evaluated patients in the supine position, while sitting on the edge of a bed (immediate, 3 and 5 min after), and while standing (immediate and 3 min after). For blood pressure measurements, a brachial cuff was placed around the left arm kept in a fixed position at heart level. HR was measured using an electrocardiogram of the same monitor. SpO~2~ was measured using a probe that attached to the right finger. OH and the cardiovascular response while the patients transitioned from sitting on the edge of the bed to standing were evaluated. Fluid balance was defined volume loss from start of anesthesia to the morning of the day of mobility.
Vasopressors included a postoperative continuous intravenous infusion of catecholamines (i.e., dopamine, dobutamine, epinephrine, and norepinephrine). We calculated the catecholamine index (CAI) \[[@CR17], [@CR18]\], as follows, wherein all doses were expressed as μg/kg/min.
CAI = (dopamine dose × 1) + (dobutamine dose × 1) + (epinephrine dose × 100) + (noradrenaline dose × 100).
A selection of perioperative anesthetics and the use of thoracic epidural anesthesia were at the discretion of the attending anesthesiologist. Postoperative analgesics use included continuous intravenous infusion of fentanyl and epidural infusion of ropivacaine hydrochloride. For intravenous infusion of fentanyl, a Terufusion syringe pump TE-331S (TERUMO, Tokyo, Japan) was used. For epidural infusion anesthetics (ropivacaine 0.2%), Infuser SV4 (Baxter Limited, Tokyo, Japan) was used. The basal infusion rate was set at 4 ml/h.
Assessment of OH {#Sec5}
----------------
OH was defined as a fall in SBP of at least 20 mmHg or a fall in DBP of at least 10 mmHg within 3 min upon standing and as intolerable dizziness, nausea and vomiting upon mobilization \[[@CR9], [@CR19]\]. Mobilization sessions were discontinued when the patients experienced an increase in symptoms until they returned to a sitting position (fainting, excessive pain, dizziness, nausea, sweating, pallor, and postoperative delirium) and when the patients could not measure at baseline vital sign.
Outcome measures {#Sec6}
----------------
The primary outcome was to clarify the incidence of OH during early postoperative mobilization after having undergone cardiothoracic and abdominal surgery and the relationship between early mobilization and cardiovascular response changes. The secondary outcome was to identify independent factors associated with OH by performing multivariate logistic regression.
Statistical analysis {#Sec7}
--------------------
Data were compared using a Wilcoxon's rank sum test and Fisher's exact test between cardiothoracic and abdominal groups. Wilcoxon's rank sum test was performed for continuous variables and Fisher's exact test was performed for categorical variables. Multivariate logistic regression analysis was then performed to correct for risk factors that showed at least a trend toward significance (*P* \< 0.1) in the univariate analysis. Multivariate logistic regression analysis was used to identify the dependent variables of orthostatic hypotension. Data are described as frequencies for categorical variables and as the median and interquartile range (IQR) for quantitative variables. *P*-values \<0.05 were considered statistically significant. Statistical analysis was performed using JMP 11.0 software (SAS Institute Japan, Tokyo, Japan).
Results {#Sec8}
=======
A total of 544 patients who had undergone cardiothoracic and abdominal surgery were evaluated in a prospective cohort study (Fig. [1](#Fig1){ref-type="fig"}). Since data of 49 patients were missing, they were excluded. Table [1](#Tab1){ref-type="table"} shows the baseline characteristics of the patients. Four hundred ninety-five postoperative patients were categorized into cardiac group (*n* = 119, aged 74 IQR, 66-79 years), thoracic group (*n* = 208, aged 65 IQR, 58-73 years), and abdominal group (*n* = 168, aged 72 IQR, 64-78 years).Fig. 1Flow of patients through the study Table 1Patient demographicsCardiac group\
(*n* = 119)Thoracic group\
(*n* = 208)Abdominal group\
(*n* = 168)Sex, male / female82 / 37140 / 68113 / 55Age, years74 (66 to 79)65 (58 to 73)72 (64 to 78)BMI, kg/m^2^18 (16 to 20)18 (16 to 20)18 (16 to 20)Comorbidities, n (%) Cardiovascular35 (29.4)50 (24.0)55 (32.7) Respiratory4 (3.4)38 (18.3)21 (12.5) Neurological5 (4.2)4 (1.9)7 (4.2) Orthopedic0 (0)3 (1.4)20 (11.9) Preoperative ejection fraction, %64 (54 to 72)69 (65 to 73)71 (65 to 75) Operative time, min254 (197 to 323)252 (186 to 390)426 (306 to 570) Operative blood loss, g700 (400 to 1228)183 (70 to 446)600 (275 to 1200) Blood transfusion, ml560 (280 to 1120)0 (0)0 (0 to 289) Fluid balance, ml280 (101 to 569)413 (148 to 680)826 (332 to 1336) Vasopressor, n (%)65 (54.6)0 (0)5 (3.0) CAI0.2 (0 to 0.4)00 Analgesia with opioids, n (%)19 (16.0)57 (27.4)141 (83.9) Epidural anesthesia, n (%)0 (0)130 (62.5)0 (0) Postoperative hemoglobin, g/dl10 (10 to 11)10 (8 to 11)11 (10 to 12) Serum creatinine, g/dl1.0 (0.7 to 1.3)0.7 (0.5 to 0.8)0.8 (0.6 to 0.9)The rate of variability of vital signs ⊿SBP, %-2 (6 to −13)−8 (1 to −18)−4 (8 to −16) ⊿DBP, %0 (8 to −11)−5 (6 to −15)0 (9 to −8) ⊿MAP, %−3 (8 to −10)−7 (3 to −15)−2 (8 to −10) ⊿HR, %4 (10 to 0)9 (15 to 2)11 (19 to 5) Initial sitting, day2 (1 to 3)1 (1 to 1)1 (1 to 1) Initial standing, day2 (1 to 4)1 (1 to 1)1 (1 to 1)*Notes*: Values were reported as the median (IQR) or number and percentage of subjects*Abbreviations*: *BMI* Body mass index, *CAI* Catecholamine index, *OH* Orthostatic hypotension, *IQR* Interquartile range, *DBP* Diastolic blood pressure, *HR* Heart rate, *MAP* Mean arterial pressure, *SBP* Systolic blood pressure; ⊿: The range of variation for cardiovascular responses
Coronary heart disease 38 (31.9%), valve diseases 57 (47.9%), aortic and vascular disease 24(20.2%) patients were included in cardiac group. Lung cancer 134 (64.4%), mediastinal neoplasm 12 (5.8%), esophageal cancer 39(18.8%), Other 23 (11.1%) patients were included in Thoracic group. Gastric cancer 33 (19.6%), colorectal cancer 9 (5.4%), liver and biliary system diseases 79 (47.0%), pancreatic cancer 47 (28.0%) patients were included in abdominal group.
Cardiovascular comorbidities included hypertension in 144 patients (29.1%), arrhythmia in 12 patients (2.4%), and angina pectoris in 5 patients (1.0%). Comorbidities of the respiratory system included chronic obstructive pulmonary disease in 12 patients (2.4%), asthma in 7 patients (1.4%), and interstitial lung disease in 5 patients (1.0%). Neurological system comorbidities included cerebrovascular disease in 5 (1.0%) patients. Orthopedic system comorbidities included vertebral compression fractures and lumbar spinal canal stenosis in 5 (1.0%) patients. Median preoperative ejection fraction in all patients was 70 (65 to 75). Figure [2](#Fig2){ref-type="fig"} shows the cardiovascular response during postoperative mobilization. There were no significant differences in the cardiovascular response between the groups. There were significant differences in type of surgery (*p* = 0.02), analgesia with opioids (*p* = 0.02) and epidural anesthesia (*p* \< 0.0001) between those with and without OH (Table [2](#Tab2){ref-type="table"}).Fig. 2Comparison of the range of variation in cardiovascular response. **a** Systolic blood pressure, {**b**) diastolic blood pressure, (**c**) mean arterial pressure, (**d**) heart rate. OH; orthostatic hypotension Table 2Demographics of patients who with or without OHWith OH (*n* = 191)Without OH (*n* = 304)*p* ValueSex, male / female138 / 53197 / 1070.08Age, years69 (61 to 77)70 (61 to 77)0.81BMI, kg/m^2^18 (16 to 20)18 (16 to 20)0.57Comorbidities, n (%) Cardiovascular47 (24.6)93 (30.6)0.15 Respiratory21 (11.0)42 (13.8)0.36 Neurological6 (3.1)10 (3.3)0.93 Orthopedic8 (4.2)15 (4.9)0.70 Preoperative ejection fraction, %67 (63 to 72)70 (65 to 75)0.06Type of surgery, n (%)0.02 Cardiac surgery39 (32.8)80 (67.2) Thoracic surgery95 (45.7)113 (54.3) Abdominal surgery57 (33.9)111 (66.1) Operative time, min302 (202 to 438)305 (2205 to 439)0.62 Operative blood loss, g360 (154 to 835)430 (148 to 805)0.89 Blood transfusion, ml0 (0 to 304)0 (0 to 520)0.73 Fluid balance, ml538 (187 to 1093)442 (125 to 910)0.18 Vasopressor, n (%)36 (18.9)44 (14.8)0.27 CAI0 (0 to 0)0 (0 to 0)0.13 Analgesia with opioids, n (%)71 (37.2)146 (48.0)0.02 Epidural anesthesia, n (%)74 (38.7)56 (18.4)\< .0001 Postoperative hemoglobin, g/dl11 (10 to 12)11 (10 to 12)0.74 Serum creatinine, g/dl0.8 (0.6 to 1.0)0.8 (0.6 to 1.0)0.42 Initial sitting, day1 (1 to 2)1 (1 to 2)0.94 Initial standing, day1 (1 to 2)1 (1 to 2)0.28*Notes*: Values were reported as the median (IQR) or number and percentage of subjects*Abbreviations*: *BMI* Body mass index, *CAI* Catecholamine index, *OH* Orthostatic hypotension, *IQR* Interquartile range, *DBP* Diastolic blood pressure, *HR* Heart rate, *MAP* Mean arterial pressure, *SBP* Systolic blood pressure
One hundred and ninety-one (38.6%) of 495 patients exhibited OH. The incidence of OH was 39 (32.8%) of 119 in cardiac patients, 95 (45.7%) of 208 in thoracic patients, and 57 (33.9%) of 168 in abdominal patients. Forty-eight (25.1%) of 191 patients withdrew during postoperative mobilization. Of these 48 patients, seven (17.9%) of 39 patients had undergone cardiac surgery, 31 (32.6%) of 95 patients had undergone thoracic surgery, and 10 (17.5%) of 57 patients had undergone abdominal surgery (Fig. [3](#Fig3){ref-type="fig"}). Early mobilization was discontinued in 48 patients due to dizziness in 16 patients (33.3%), nausea in 15 patients (31.3%), wound pain in 9 patients (18.6%), and fatigue in 8 patients (16.7%). There was no significant correlation between discontinuation and the rate of variability in SBP (*p* = 0.74), DBP (*p* = 0.28), HR (*p* = 0.50) and MAP (*p* = 0.52).Fig. 3Incidence rate of OH and rate of discontinuation during postoperative mobilization
There were significant differences in blood transfusion (*p* = 0.01), vasopressor (*p* = 0.02) and CAI (*p* = 0.01) between those with and without OH in the cardiac group. There were significant differences in fluid balance (*p* = 0.01), analgesia with opioids (*p* \< 0.05) and epidural anesthesia (*p* \< 0.001) between those with and without OH in the thoracic group. There were significant differences in operative time (*p* = 0.02), operative blood loss (*p* = 0.01), blood transfusion (*p* \< 0.05), vasopressor (*p* \< 0.05), and initial standing (*p* = 0.02) between those with and without OH in the abdominal group (Table [3](#Tab3){ref-type="table"}).Table 3Demographics of patients in type of surgery who with or without OHCardiac groupThoracic groupAbdominal groupWith OHWithout OH*p*-valueWith OHWithout OH*p*-valueWith OHWithout OH*p*-value(*n* = 39)(*n* = 80)(*n* = 95)(*n* = 113)(*n* = 57)(*n* = 111)Sex, male / female8 / 3151 / 290.0868 / 2772 / 410.2339 / 1874 / 390.82Age, years72 (61 to 78)75 (69 to 79)0.0667 (60 to 73)64 (57 to 72)0.0872 (67 to 78)71 (63 to 78)0.45BMI, kg/m^2^18 (17 to 19)18 (16 to 20)0.7218 (16 to 20)18 (16 to 20)0.9718 (15 to 19)18 (16 to 20)0.50Comorbidities, n (%) Cardiovascular12 (31.0)27 (33.8)0.8217 (17.9)33 (29.2)0.0718 (31.6)37 (33.3)0.86 Respiratory04 (5.0)0.0714 (14.7)24 (21.2)0.237 (12.3)14 (12.6)0.95 Neurological1 (2.6)4 (5.0)0.522 (2.1)2 (1.8)0.863 (5.3)4 (3.6)0.69 Orthopedic002 (2.1)1 (0.9)0.677 (12.3)13 (11.7)0.91 Preoperative ejection fraction, %63 (57 to 70)67 (53 to 73)0.5367 (62 to 73)71 (63 to 75)0.1465 (60 to 69)68 (62 to 74)0.07 Operative time, min227 (190 to 316)263 (199 to 331)0.14252 (192 to 318)252 (180 to 486)0.52492 (366 to 570)390 (264 to 564)0.02 Operative blood loss, g575 (285 to 848)750 (470 to 1420)0.06180 (85 to 460)189 (52 to 405)0.38870 (330 to 1810)515 (181 to 915)0.01 Blood transfusion, ml350 (188 to 708)730 (280 to 1150)0.010 (0 to 0)0 (0 to 0)0.400 (0 to 560)0 (0 to 0)\< 0.05 Fluid balance, ml591 (246 to 1267)607 (242 to 1104)0.77592 (299 to 1218)441 (109 to 1007)0.01347 (104 to 650)322 (111 to 631)0.70 Vasopressor, n (%)27 (69.2)38 (47.5)0.025 (5.3)5 (4.4)0.784 (7.0)1 (0.9)\< 0.05 CAI0.3 (0 to 0.5)0 (0 to 0.3)0.010 (0 to 0)0 (0 to 0)0.770 (0 to 0)0 (0 to 0)-- Analgesia with opioids, *n* (%)6 (15.4)13 (16.3)0.9016 (16.8)41 (36.3)\< 0.0549 (86.0)94 (84.7)0.83 Epidural anesthesia, n (%)000 (0 to 0)56 (49.6)\< 0.00100 Postoperative hemoglobin, g/dl10.1 (9.5 to 11.1)10.3 (9.6 to 11.6)0.4111.7 (10.7 to 12.8)11.9 (10.8 to 13.1)0.2110.7 (10.0 to 11.4)10.8 (8.8 to 14.4)0.94 Serum creatinine, g/dl0.9 (0.7 to 1.4)1.0 (0.7 to 1.3)0.770.7 (0.5 to 0.8)0.7 (0.6 to 0.8)0.930.8 (0.6 to 1.0)0.8 (0.6 to 0.9)0.83 Initial sitting, day2 (1 - 4)2 (1 - 3)0.871 (1 - 1)1 (1 - 1)0.671 (1 - 1)1 (1 - 2)0.06 Initial standing, day2 (2 - 4)2 (1 - 3)0.371 (1 - 1)1 (1 - 1)0.351 (1 - 1)1 (1 - 2)0.02*Notes*: Values were reported as the median (IQR) or number and percentage of subjects*Abbreviations*: *BMI* Body mass index, *CAI* Catecholamine index, *OH* Orthostatic hypotension, *IQR* Interquartile range
According to the postoperative medication received, antihypertensive agents were prescribed to 29 (24.4%) patients, antiarrhythmic agents to 24 (20.2%) patients in cardiac group. These medications were no significant differences in the cardiovascular response between the groups. Nineteen (16.0%) of 119 patients in the cardiac group received postoperative continuous intravenous infusion of fentanyl, with the median rate of infusion being 25 μg/h (IQR, 25-50 μg/h). Of the 208 thoracic surgery patients, 130 (62.5%) patients received a continuous postoperative epidural infusion of ropivacaine, with the median rate of infusion being 8 μg/h, and 57 (27.4%) received fentanyl, with the median rate of infusion being 50 μg/h (IQR, 50-50 μg/h). Of the 168 abdominal patients, 141 (83.9%) received fentanyl with the median rate of infusion being 50 μg/h (IQR, 28-50 μg/h).
In a univariate analysis of the following parameters, the type of surgery, sex, age, body mass index, preoperative ejection fraction, fluid balance, operative time, blood loss, the value of hemoglobin and serum creatinine measured on POD 1, CAI, the use of vasopressors, analgesia with opioids, and epidural anesthesia showed a significant relationship with OH.
Multivariate logistic regression analysis indicated an independent association between the incidence of OH and male sex (Odds ratio (OR); 1.538, 95% CI, 1.027 to 2.326; *p* = 0.03) or receiving epidural anesthesia (OR; 2.906, 95% CI, 1.924 to 4.419; *p* \< 0.001; Table [4](#Tab4){ref-type="table"}).Table 4Results for determining factors of OHVariableUnadjustedAdjustedOR(95% CI)*p* ValueOR(95% CI)*p* ValueAge0.10(0.966 to 1.026)0.83BMI1.08(0.974 to 1.207)0.15Sex (Male = 1, Female = 0)2.346(1.170 to 4.705)0.011.538(1.027 to 2.326)0.03Type of surgery Thoracic surgery1.0 \[Reference\] Cardiac surgery1.491(0.453 to 4.909)0.59 Abdominal surgery1.933(0.810 to 4.615)0.14Preoperative ejection fraction1.019(0.992 to 1.050)0.21Operative time0.984(0.875 to 1.109)0.79Operative blood loss1.000(1.000 to 1.000)0.97Fluid balance, ml0.100(1.000 to 1.000)0.85Vasopressor (with = 1, without = 0)0.425(0.151 to 1.197)0.53Analgesia with opioids (with = 1, without = 0)1.433(0.615 to 3.338)0.40Epidural anesthesia (with = 1, without = 0)4.329(1.064 to 17.617)0.042.906(1.924 to 4.419)\< 0.001Postoperative hemoglobin0.917(0.735 to 1.140)0.41*Abbreviations*: 95% CI: 95% confidence interval, *BMI* Body mass index, *CAI* Catecholamine index, *OR* Odds ratio, *OH* Orthostatic hypotension
Discussion {#Sec9}
==========
The main findings of the present study in patients who had undergone cardiothoracic and abdominal surgery, we found that (1) 38.6% experienced OH during early mobilization after surgery and (2) men and patients who received an epidural anesthesia were predisposed to experience OH during early mobilization.
The results suggest that the range of variation in the cardiovascular responses during postoperative mobilization is limited, and we did not recognize an association between the cardiovascular response and symptoms. In general, postoperative patients are particularly vulnerable upon mobilization aggravate the postural reduction in central blood volume in the upright position \[[@CR20]\]. However, a previous study also indicated that there was no relationship between postoperative symptoms and cardiovascular response during early mobilization \[[@CR21]\]. Our results were consistent with those of the previous report. Therefore, it is important to assess the subjective symptoms of patients in addition to measuring cardiovascular response during mobilization.
The incidence of OH and the use of postoperative opioids were higher than in previous studies \[[@CR12]--[@CR16]\], and a quarter of patients discontinued from the study. Approximately 60% of cases that discontinued early mobilization were due to dizziness and nausea. In general, the use of postoperative opioids is associated with an increased risk of nausea and vomiting, thus resulting in an increase in the incidence and severity of OH \[[@CR15], [@CR22]\]. Although it is reasonable to assume that both epidural analgesia and analgesia with opioids are the causes of OH, we found that only epidural anesthesia was associated with an increased risk of OH. The use of epidural anesthetics theoretically induces OH through sympathectomy-induced vasodilation \[[@CR16]\]. Gramigni E et al. \[[@CR23]\] reported cardiovascular response during postoperative mobilization that involved the use of thoracic epidural analgesia with a mixture of bupivacaine and fentanyl. They concluded that epidural analgesia was associated with arterial hypotension during the postoperative period. Although we agree with their opinion, our results identified that epidural analgesia without opioids was an independent predisposing factor for OH. However, it is difficult to be certain of the relationship between epidural analgesia and OH based upon the results of this study only. Overall, when employing early mobilization, clinicians and physiotherapists should be careful when postoperative patients use analgesics to treat OH. As discussed above, many studies offer different perspectives; however, it is approximately certain that analgesics influence the incidence of OH. Since most patients use analgesics after surgery, clinicians should take precautions when mobilizing these patients. Furthermore, physiotherapists should advise patients to be in Fowler's position and wear elastic stockings during the day to reduce the risk of OH \[[@CR24]\]. Early mobilization using various devices is important for prevention of orthostatic intolerance, deep vein thrombosis and pulmonary complications.
In the cardiac group, rigorous postoperative management was required to prevent heart failure and postoperative bleeding such as open aortic surgery and heart valve surgery. In this study, 55% of vasopressor use occurred in the cardiac group. When we evaluated the rate of OH and discontinuation in the cardiac group, the use of vasopressor might not have influenced the relationship between orthostatic cardiovascular responses, although vasopressor use and the CAI were significantly different between the groups with and without OH. It is reasonable for us to anticipate that the incidence of OH would be increasing without vasopressor use; however, in our results, OH was more prevalent in patients using vasopressors. Therefore, it was difficult to assess the influence of vasopressor use in the cardiac surgery group. Also, antihypertensive and antiarrhythmic agents were used in cardiac group. Although, these medicines were no significant differences in the cardiovascular response between the groups, we should continue to evaluate the relationship between these medicines use and OH development.
OH was observed more frequently in the thoracic group. Epidural anesthesia was shown to be an independent predisposing factor for OH development in multivariate analysis, and it was used only in this group. Mizota et al. \[[@CR16]\] found that postoperative opioid use were independent risk factors for OH after VATS. In addition, they demonstrated that continuous postoperative epidural administration of ropivacaine at 0.2% at a rate of 2-6 ml/h did not induce clinically significant vasodilation. In our study, we applied epidural anesthesia using a continuous postoperative epidural administration of the same dose of ropivacaine. In doing so, we found that its use was an independent predisposing factor for OH. Thus, we observed a different result from that of the previous study. Therefore, in the future, we will attempt to clarify the influence of epidural anesthesia on the development of OH.
In the abdominal group, operative factors might not be associated with OH between patients with and without OH. Haines KJ et al. \[[@CR25]\] concluded that 52% of patients undergone high-risk abdominal surgery had a barrier to mobilization, with the most common barrier being hypotension. In our study, we did not detect a relationship between blood pressure and OH. It was difficult to anticipate OH using only the cardiovascular response in patients who had undergone abdominal surgery. Future studies should include physiological parameters to predict the incidence of OH in patients who had undergone abdominal surgery.
The male was independently associated with the incidence of OH. Convertino \[[@CR26]\] demonstrated that fluid volume shifting during postural changes had been lower in females than in males. Females are at an increased risk for OH as a result of common variables related to body size and hormones \[[@CR27]\]. However, our results did not show similar results, which may possibly be due to two reasons. First, our subjects were postoperative patients. Surgical trauma imposes increased demands on organs with increased sympathetic tone and a subsequent endocrine metabolic response. In addition, blood loss and fluid volume shifts that might occur might influence the cardiovascular response to mobilization \[[@CR20]\]. Second, in our subjects, more males were enrolled than females. These differences may also be related to muscle volume and body size. Therefore, our results suggest that male postoperative patients might experience orthostatic hypotension due to increased muscle sympathetic and peripheral blood vessel outflow due to muscle contraction. Currently, sex differences in OH have been observed, but the mechanisms are still unknown \[[@CR27]\].
This study had several limitations. First, we investigated different types of surgeries because we wanted to capture the characteristics of each surgery. The mechanism of fluid shifts of each groups are probably heterogeneous due to complicated and diverse group of patients. However, the reason for targeting patients at postoperative course of thoracic and abdominal surgery is justified because these surgeries have an influence on autonomic nervous system and OH of postoperative complications are most likely to occur. Therefore, it was considered important to assess the cardiovascular response and orthostatic intolerance during mobilization after these postoperative surgeries. Second, we evaluated the phenomenon of hypotension under a clinical setting. Anatomical physiological details were unknown. Further studies are necessary to clarify these limitations.
Conclusion {#Sec10}
==========
The present study found a high incidence of OH during early mobilization in patients who had undergone cardiothoracic and abdominal surgery. Two independent predisposing factors for OH (male sex, and epidural anesthesia in thoracic surgery) were identified. In conclusion, it should be recognized that OH occurs frequently during postoperative early mobilization.
CAI
: Catecholamine index
DBP
: Diastolic blood pressure
ERAS
: Enhanced recovery after surgery
HR
: Heart rate
IQR
: Interquartile range
MAP
: Mean arterial pressure
OH
: Orthostatic hypotension
POD
: Postoperative day
PPCs
: Postoperative pulmonary complications
RR
: Respiratory rate
SBP
: Systolic blood pressure
SpO2
: Oxygen saturation
VATS
: Video-assisted thoracic surgery
The authors would like to thank Drs. Terumitsu Sawai and Takako Tanaka, from Unit of Rehabilitation Sciences, Graduate School of Biomedical Sciences, Nagasaki University, for study concept and design advice. The authors thank all of the staff who acquired data for study: Yudai Yano; Department of Rehabilitation Medicine, Nagasaki University Hospital: Naoki Mio; Department of Rehabilitation Medicine, Hiroshima University Hospital: Issei Natsui; Department of Rehabilitation Medicine, Nagasaki Harbor Medical Center City Hospital: Masaki Oomagari, Kyohei Ito, Takeo Kimura at Department of Rehabilitation Medicine, Seirei Mikatahara General Hospital.
Funding {#FPar1}
=======
None.
Availability of data and materials {#FPar2}
==================================
The authors do not wish to share our data, because the patients' who participated this study did not agree to share their individual data.
MH, YT, YM, MO and RK planned the study. MH, YM, MO collected the data. MH, TM, SS performed the statistical analysis and drafted the manuscript. MH, HN, KE, TN, SE, RK convinced of the study, and participated in its design and coordination. All authors read and approved the final manuscript.
Ethics approval and consent to participate {#FPar3}
==========================================
The Human Ethics Review Committee of Nagasaki University Hospital (Approval number: 09022760-2) and Seirei Mikatahara General Hospital (Approval number: 09-02) approved this study. Participants gave written informed consent for analysis and publication prior to the study scenario. This study was registered in UMIN-CTR (UMIN000018632).
Consent for publication {#FPar4}
=======================
All authors have given final approval of the version to be published.
Competing interests {#FPar5}
===================
The authors declare that they have no competing interests.
Publisher's Note {#FPar6}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
Plain English summary {#Sec1}
=====================
Although scientific and technological progress has provided major health benefits, it can also hamper communication between healthcare personnel and patients. In modern medicine, the technical aspects of healthcare tend to prevail, in which the main focus is placed on treating the disease and less attention is paid to other aspects that are also important to the patient, such as feelings and emotions. Knowing how to communicate is an ethical and legal imperative and therefore healthcare professionals must ensure that patients are aware of everything related to their condition, to facilitate their autonomy in decision-making. If knowing how to communicate is always important, it is even more so when the content of the message is unfavourable. For example, bad news about the advance of a pregnancy can influence the mother's decision on whether to continue or to interrupt it. Professional interventions in such cases are crucial, because the psychological consequences of the situation depend on the care and support provided. Deficiencies in the communication process can generate conflicts and dissatisfaction in the professional-patient-family relationship.
Background {#Sec2}
==========
Technological progress and advances in the field of molecular genetics have contributed decisively to the development of prenatal diagnostics. Biotechnology offers novel, highly reliable instruments and techniques for identifying maternal risk factors, enabling the early detection of congenital malformations or defects of diverse types related to foetal formation and development \[[@CR1]\]. However, despite this considerable progress, there is often a profound lack of interest in interpersonal communication. The ability to communicate with patients has been erroneously viewed as a lesser skill, compared to technical aspects of healthcare \[[@CR2], [@CR3]\], when in fact it is an essential element in the relationship between healthcare personnel and their patients, based on mutual recognition and shared decision making \[[@CR4]--[@CR6]\].
When a chromosome defect or severe foetal malformation is confirmed, the health team is obliged to inform the mother about the advisability of continuing the pregnancy, and of the prognosis and possible postpartum outcomes \[[@CR7]--[@CR10]\] respecting the patient's right to choose \[[@CR4]\]. In many cases, healthcare personnel develop their own strategies, without taking into account the patient's holistic nature and without forming a comprehensive outlook on the health-disease binomial, capable of transforming information into a therapeutic tool \[[@CR11]\]. Carers tend not to listen, thereby failing to obtain feedback to clarify exactly what the patient requires, and becoming emotionally distant. This inability to adapt the communication of diagnostic information to the patient's own values and preferences can generate conflicts and dissatisfaction \[[@CR11]--[@CR14]\]. Attitudes and communicative skills play a fundamental and decisive role in addressing problems, helping patients overcome psychological distress and conveying realistic expectations \[[@CR15]\]. In prenatal preventive care programmes, the evaluation of obstetric risk, as well as ruling out biologically-based problems and identifying women at high risk of maternal and perinatal complications, should also address the patient from an inclusive perspective, because (among other reasons) the health-disease process is a multidimensional one in which biological, psychological and social factors continuously interact, in a positive or a negative sense \[[@CR16]\].
The aim of this study is to examine the healthcare provided to pregnant women whose foetuses present congenital defects and to facilitate the design of a more personalised health model, one that responds to all their needs, rather than a limited selection. Accordingly, we have analysed the views expressed by healthcare personnel and the interventions they make when an adverse prenatal diagnosis must be communicated, including the strategies and skills employed to meet patients' information needs and to respond to their emotional responses.
Methods {#Sec3}
=======
This qualitative study was undertaken from a phenomenological standpoint, which enabled us to analyse specific experiences \[[@CR17]\]. Information was collected by means of non-participant observation and semi-structured interviews, obtaining opinions from physicians, midwives, nurses and nursing assistants involved in real-life care processes relevant to the object of our study. Our analysis is based on descriptive and interpretative paradigms, which were applied to interpret the data obtained in the context of existing knowledge about the study area and in that of the participants' experiences in this respect. To maximise the scientific and methodological rigour of the study, criteria of credibility, transferability, dependence (or consistency) and confirmability were applied \[[@CR17], [@CR18]\].
The study was carried out at the Costa del Sol Health Agency (Marbella, Spain) from June to September 2015. During the information-compiling phase, 37 interviews were conducted, with 22 obstetricians, four midwives, three nurses and eight nursing assistants (Table [1](#Tab1){ref-type="table"}). These healthcare personnel all worked in units within the Obstetrics service (prenatal diagnosis, perinatal medicine, hospitalisation, first, second and third-trimester pregnancy consultation and/or obstetric ambulatory care). The selection of the participants was intentional and not random, taking into account the criterion of segmentation, depending on the occupational category, hospital unit, position occupied and duration of employment at the hospital. The inclusion criteria were that participants should be physicians, midwives, nurses or nursing assistants at the hospital, have at least 1 year's experience (to acquire sufficient knowledge of the unit) and give permission for audio recording of the interview. All potential participants were sent letters indicating the purpose of the study and inviting them to take part. Participation was later confirmed by telephone.Table 1Sociodemographic characteristics of healthcare professionalsProfessional categoryObstetricianMidwifeNurseNursing assistantSex Male Age \<4021 41--45 46--50 \>502 Female Age \<4011223 41--45512 46--50211 \>502Total22438Mean age (Years)39384044Mean Experience (Years)11101624
The principal investigator[1](#Fn1){ref-type="fn"} directly observed the interactions between patients and healthcare personnel, paying attention both to their words and to non-verbal aspects. A field diary was used to make a detailed record of all observations, after patients were informed and gave their written consent to the procedure. To avoid subjectivity, a feedback process was also applied, in which the notes made and impressions received were later shared with the same participants during the interviews.
The script for the interviews was elaborated ad hoc, taking into account our previous review of the literature and the dimensions of the study, and was examined and agreed upon by all members of the research team. The recordings were transcribed and NVivo 11 qualitative software was used to encode the information and to perform the content analysis. The qualitative data analysis was carried out using the Taylor-Bogdan system, based on data preparation, identification of emerging issues, coding, interpretation, relativisation and determination of methodological rigour \[[@CR19], [@CR20]\].
Both emerging issues (those arising during the interview) and predefined ones (discussion topics included in the interview design) were identified. To ensure optimum data quality, triangulation was applied, regarding both the data (using data compilation instruments such as interviews and non-participant observation) and the researchers (the analysis was performed by two researchers^1,3^, first working independently and then reaching a consensus view) \[[@CR21]\].
Data saturation was determined after analysing the number of references encoded and dimensions identified, as the point at which the reading and coding process failed to generate additional information that would require further codes or categories.
Results {#Sec4}
=======
Data analysis revealed the existence of three related categories: how to give bad news, communication skills in general and interactions between healthcare professionals and patients.
The following results are presented in terms of the corresponding information category, highlighting the main findings obtained for each subcategory. These findings are also shown in the tables, together with verbatim transcripts from the interviews. The occupation of the healthcare professional is identified by a code assigned to the interview, thus ensuring anonymity and confidentiality. Finally, the data obtained are considered in the light of previous research reports in this regard.
How to give bad news {#Sec5}
--------------------
### Issuers and receivers of bad news {#Sec6}
Healthcare workers emphasise the unexpected nature of bad news and the unwanted changes they provoke in the lives of those who receive it.
Regarding who should perform this communication, there is general agreement that this should be the obstetrician, without excluding the support that other team members may provide, as long as they are well informed about the case. All accept that the pregnant woman should be the first to receive the message, that her autonomy as a human being should be recognised and that she has the right to be informed about everything related to her own health and to that of the foetus. Obstetricians stress how important it is to inform the woman's partner, too, from the outset or, if this is impossible, a person of trust who can provide support and liaison between the parties.
With respect to the medical consultation, the general opinion of the participants is that the person accompanying the mother (usually, the partner or a close relative) should have the patient's consent to hear the details about the evolution of pregnancy, as part of the clinical relationship. None of the healthcare workers believe that information should be concealed or that any pact of silence should be made with the family.
### What bad news is communicated and how {#Sec7}
Among the most frequent areas of bad news mentioned are diagnoses of miscarriage, chromosomal alteration and intrauterine foetal malformation or death. According to the participants, the early detection of congenital anomalies, in prenatal diagnosis, usually means that a decision must be taken on whether or not to interrupt the pregnancy.
Regarding how it is communicated, from the standpoint of the healthcare professionals, it is advisable to give immediate notice of the discovery of bad news and to inform the patient of the alternatives available. Whatever the prognosis, the availability of equipment to control the symptoms and to ensure the best possible quality of life and comfort should be stressed.
Most of the participants agreed that information should be transmitted gradually, and if possible in line with the patient's wish to receive it. According to healthcare workers, the patient's first reaction is often to suffer a state of shock, in which she is unable to process the information provided. A nursing assistant who works in pregnancy monitoring consultations and who, moreover, has a degree in psychology, illustrates this very well with the following comment:
### AUXE01 {#Sec8}
*"Gradually. I think that we should tell them everything, that the information must be complete, and not concealed or ignored, no matter how painful it may be. On the other hand, it mustn't be blurted out too abruptly. All extremes are bad; being told little by little might be agonising, but all at once can be devastating. It's important to enable feedback; you tell them the situation and, depending on the reactions you get, you know where to place more emphasis and where you need to tread lightly."*
Regarding how the information should be provided, some professionals choose to prepare the ground before entering fully into the question, others prefer to convey optimism and to leave a door open to hope during the first contact with the patient. In any case, for those interviewed the communication of bad news is not a "single or isolated act" that ends here and now, but must be viewed as a process that requires time and effort. The healthcare team believes it important to inform the person affected that there will be more meetings and opportunities for her to express doubts that may arise, or simply her pain and anger.
The healthcare professionals consulted in our study consider it important to avoid the excessive use of technical terms and to adapt their verbal language to the cultural level of the patient, as detected at the time of the interview. Among the notes made in field diaries during the observation, it is significant that negative reactions were aroused among patients by certain expressions commonly used in obstetrics terminology, such as miscarriage, malformation or foetal death. In this respect, opinions among the professionals are divided: the majority consider that such terms must be used, in order to avoid ambiguity, but others prefer to use alternative expressions, in order to soften the emotional impact produced.
Within this subcategory, an aspect of some importance is that of non-verbal language in interactions with the patient. The researchers noted a reaction that was also observed by many of the obstetricians consulted, namely that the non-verbal language used by healthcare professionals during the ultrasound exploration produces expectation and concern in the women being examined.
### Optimum environmental conditions when bad news must be given {#Sec9}
Noise and interruptions are the main barriers to establishing a climate of trust during meetings with patients. The healthcare workers commented that they often have to improvise a space in which to hold these meetings, since no purpose-designed physical location is reserved for this type of communication.
These professionals agree that the ideal place for the communication of bad news should provide privacy and tranquillity, be separated from the maternity area, be comfortable and have enough natural light.
### The evolution of carer-patient communication {#Sec10}
According to the professionals interviewed, scientific-technological advances, the increasing presence of women in obstetrics and gynaecology teams and the practice of defensive medicine are the main factors underlying the evolution of communication and clinical relationships with patients.
### The role of nursing staff {#Sec11}
The obstetricians consulted acknowledge the commendable work done by nurses and auxiliaries in prenatal consultations, in perinatal medicine and during the hospital stay of women who decide to terminate their pregnancy. The support offered by these hospital workers is viewed as an essential element in achieving patients' satisfaction and collaboration.
### Strategy and summary {#Sec12}
Many healthcare workers professionals commented that when significant information must be transmitted patients should be given sufficient time to relax and to consider the situation and any doubts they may have. Accordingly, they are asked to leave the consultation during this time, to allow other patients to be attended. If this period of reflection is considerable, the patient may be offered an appointment for another day, especially if crucial decisions must be taken.
According to healthcare staff, lack of time (often caused by an overload of responsibilities) is responsible for the absence of feedback on the quality of the attention provided.
### Setting of the intervention {#Sec13}
Sometimes the suspicion or the discovery of foetal malformation arises during a routine ultrasound exploration, or the patient may be asked to attend a consultation to be given the confirmation of a previous finding. For obstetricians, the presence or absence of prior knowledge of the diagnosis and the time available are the factors that will determine the preparation or improvisation of their discourse. In this respect, healthcare workers also recognise that taking control of the situation is more complicated and that anxiety and unrest may be provoked when the diagnosis arises unexpectedly (Table [2](#Tab2){ref-type="table"}).Table 2Category.- How to give bad newsSubcategory: issuers and receivers of bad news\
Verbatim***OBST S4*** *: "A woman who is pregnant always expects everything to go well, that the delivery will go well and that a healthy baby will be born; anything that goes wrong is bad news".****MID S2*** *: "Unexpected information that causes sadness and pain and provokes a change in your life".****MID S1*** *: "Something that doesn't fit what you had expected, that spoils plans and means you have to adapt to a new situation".****OBST S17*** *: "The doctor is the person who can and should communicate bad news".****NASST S8*** *: "The news has to be given by the doctor and we are there to provide support".****MID S4*** *: "Ideally, there should be a multidisciplinary team, including nurses and, if possible, a psychotherapist".****NUR S3*** *: "Although the doctor gives the news, nurses are also involved, because the patients ask us to clarify what they don't understand".****OBST S9*** *: "Obviously, you have to tell the patient and her partner, if there is one, and then assess the advisability of informing the family".****OBST S2*** *: "According to the rules on patient autonomy, the patient must always be told, and we assume that whoever is with her can also receive the news".****NUR S2*** *: "The patient must be informed because she is the one who is pregnant".****MID S2*** *: "It is very important that the information should also be received by someone the patient trusts (...)"*Subcategory: what bad news is communicated and how\
Verbatim***NASST S6*** *: "(\...) miscarriages, foetal malformations, syndromes, foetal heart disease, proposal to interrupt the pregnancy".****NUR S3*** *: "In the ward, we have patients who decide to abort because of a diagnosis of severe malformation or chromosomal alteration".****MID S12*** *: "Very often, non-viable pregnancies, anomalies detected by ultrasound scan, antepartum deaths, intrauterine growth problems".****OBST S8*** *: "There is no magic formula. Usually the news is given little by little so that the information can be assimilated, but this does not always work, and for some patients it is very painful. Although for others, telling it all at once can be devastating (\...)"****OBST S13*** *: "(\...) using the same means seen to be successful when done by more experienced colleagues".****MID S3*** *: "You create your own style, by doing it over and over again".****OBST S13*** *: "(\...) I prefer to be totally frank. Sometimes, only when you say that the foetus is dead or that its situation is incompatible with life does the patient realise the gravity of the news".****NUR S3*** *: "We try to take into account the patient's socio-cultural level, but we often forget and use too much medical jargon".****NASST S4*** *: "(\...) we aren't very close to the patients, we don't take their hands, we don't give them a hug, we avoid looking directly into their eyes, we focus on filling in the report, on the computer and on the ultrasound scan".*Subcategory: optimum environmental conditions when bad news must begiven\
Verbatim***NUR S1*** *: "Somewhere private, without interruptions, separated from the maternity area, comfortable and with sufficient natural light".****OBST S13*** *: "(\...) knocking at the door, people coming in and out, telephones ringing continually".****MID S2*** *: "The intentions are good, but there is no area specially equipped for this purpose. We try to assign a single room for a pregnancy termination, but it is not always possible".*Subcategory: the evolution of carer-patient communication\
Verbatim***OBST S8*** *: "We always have to bear in mind the question of the medical professional's legal defence. The social situation makes this inevitable, but it makes it very difficult to provide personalised, direct treatment".****NUR S2*** *: "Consent forms, signatures in duplicate, the next appointment with another healthcare professional \... all of this greatly interferes with the doctor-patient relationship".****OBST S6*** *: "We pay more attention to the diagnostics, we've gained in technological capabilities and lost in human quality".*Subcategory: the role of nursing staff, according to the physician\
Verbatim***OBST S11*** *: "The nurses provide very important support; they help us convey the message we want the patient to receive".****OBST S1*** *: "They can provide support, but diagnosis is the doctor's job and we have to communicate the message, even if we don't like it".****OBST S12*** *: "(\...) nurses spend many hours at the patient's bedside, so they are well aware of the patient's fears and expectations".*Subcategory: strategy and summary\
Verbatim***OBST S16*** *: "The extra time you give to one patient is time you're taking away from another. Ideally, an appointment should be made for another day".****NASST S7*** *: "Patients are seen again at the end of the consultation, to answer their questions".****OBST S14*** *: "At the end of the interview I do not have enough time for the patient to repeat everything I have explained to her and check if she has understood me".*Subcategory: setting of the intervention\
Verbatim***OBST S10*** *: "With experience, you usually know what to do, but patients can ask unpredictable questions that you have to address on the spot".****OBST 16*** *: "Prior awareness or otherwise of the diagnosis and of the time available determines whether the talk to the patient can be prepared or must be improvised".****OBST S13*** *: "Inadequate training in communication is a handicap, making it hard to adapt the discourse to meet all the patient's needs".OBST* Obstetrician, *MID* Midwife, *NUR* Nurse, *NASST* Nursing assistant
Communication skills {#Sec14}
--------------------
### Training in the communication of bad news {#Sec15}
The healthcare personnel who took part in this study, especially the obstetricians, commented that their academic training in communication and counselling was practically non-existent. They also observed that their acquisition of this type of knowledge was based solely on participation in a short course during initial training. Neither is specific postgraduate training provided; thus, most professionals define themselves as self-taught, having observed and learned from actions seen to be useful for other colleagues. However, these workers consider it extremely important to learn strategies that foster the creation of a solid and, above all, therapeutic clinical relationship.
Most of these professionals admit that a lack of communication techniques and social skills sometimes creates a barrier that can impede the development of a good clinical relationship. The kind of training that they consider appropriate to alleviate this problem, helping them to establish effective communication with patients and family members, would be based on periodic role-playing workshops, debriefing sessions and counselling by a psychologist.
The characteristics or skills that professionals believe should be possessed by those responsible for communicating bad news are proximity, empathy and assuredness.
### The ability to explore psycho-social issues {#Sec16}
Concerning the psychological aspect, we examined the extent to which healthcare professionals explore the patient's state of mind before communicating bad news. The hypothetical case was raised that the patient was experiencing considerable stress (the death of a loved one, or a recent serious diagnosis, either personal or affecting a close family member).
Within the social sphere, we considered how professionals explore the impact of bad news on patients' daily lives and on their families, social circles and work environments.
In response, the healthcare workers commented that no formal exploration is usually made of psychological questions or of the patient's social sphere, in terms of a systematic examination, and emphasised that this type of inquiry does not influence the communication of bad news. Although such an investigation might be undertaken in the private sphere, these professionals consider it very difficult to do so in the public domain since the way in which services are structured and consultations planned means that there is insufficient time to conduct a formal, regulated study.
The healthcare professionals also stated that patients' psychological and social problems should be treated in primary care and then, if appropriate, referred to a mental health clinic or to social workers.
### Responding to the patient's emotions {#Sec17}
We also inquired how healthcare professionals address the emotional responses of patients who are given bad news. In general, these workers observed that providing resources to help patients adapt to the new situation, alleviate their mental pain and reduce stress, depression or anxiety is outside their field of competence as gynaecologists and obstetricians, an area in which they lack training, and therefore that this task would be undertaken more appropriately by a psychotherapist.
The professionals who took part in our study did not consider themselves well acquainted with counselling strategies or resilience models for the effective management of patients' emotions.
Finally, the possible existence of language barriers was attributed to the cultural diversity that characterises the population of pregnant women attended at the hospital where this study was performed (Table [3](#Tab3){ref-type="table"}).Table 3Category.- Communication skillsSubcategory: training in the communication of bad news\
Verbatim***OBST S14*** *: "In the 2nd year at university, we did workshops on doctor-patient communication, but little else in the 6 years spent in the faculty."****OBST S12*** *: "During the first year of residency, we had a course on the doctor-patient relationship, in which we discussed the quality of care and the communication of bad news. But nothing since then."****NUR S3*** *: "(\...) I've had training in helping and in the humanisation of care, but that was a few years ago."****MID S3*** *: "Every year we offer a course on how to respond to perinatal grieving, but hardly any of the medical staff take it."****OBST S13*** *: "We would need to learn communicative techniques and skills through role playing and recordings of our own interventions, and then analyse them."****MOBST S1*** *: "(\...) we need advice from a psychotherapist, and medical team sessions to make our criteria consistent."*➣Desirable qualities in the person who must transmit bad news:***OBST S11*** *: "Sensitivity and humanity, I think."****OBST S12*** *: "Empathy with the patient and showing self-assuredness in what you have to convey."****OBST S6*** *: "Having sufficient knowledge of pathology, of what can and can't be done, and time in which to carry out possible solutions."****NASST S1*** *: "Closeness, putting yourself in the patient's place and speaking in terms that she can understand."*Subcategory: the ability to explore psycho-social issues\
Verbatim***ASST S7*** *: "This isn't examined. Some patients will say they've had a stressful experience, but the doctor doesn't go into this question, there isn't time."****OBST S2*** *: "(\...) that isn't examined. I honestly don't know what kind of inquiry might be made. If the patient has problems of this type, she usually tells you herself."****MID S2*** *: "The psychological and social aspects aren't considered due to our feelings of insecurity. We make the excuse that we don't have time, but it depends to a great extent on each individual's attitude and personal interest in the matter."****MID S4*** *: "Unless the patient tells you spontaneously (although you might intuitively sense it), you don't usually go into these areas, you only address the physical side."*Subcategory: responding to the patient's emotions\
Verbatim***NASST S1*** *: "You don't have the knowledge or skills to deal with certain problems and the easiest thing to do is to avoid them. Without specific and continuous training in the necessary areas, we can't offer patients comprehensive quality care."****OBST S3*** *: "We don't have time. To respond properly we'd need a specialised consultation, with the presence of a psychologist."*➣Counselling strategies and Models of resilience:***OBST S4*** *: "\... yes, I'd recommend it, but to help in all these areas, right now I don't have the tools, nor do we schedule appointments to assess the patient's evolution, how she's coping with the bad news or accepting it."****OBST S1*** *: "I don't know what these strategies consist of."****MAT S4*** *: "What we do is listening, and little else. The patients go home, basically, with nothing."*➣*Language barriers:****OBST S2*** *: "There are language barriers, especially with the Chinese and Arab populations, and this makes you anxious."OBST* Obstetrician, *MID* Midwife, *NUR* Nurse, *NASST* Nursing assistant
Professional-patient interaction {#Sec18}
--------------------------------
### Profile of the patients {#Sec19}
The heterogeneity of the patient population at our hospital, in terms of sociocultural status and nationality, requires healthcare professionals to be especially sensitive to the need to provide culturally appropriate care. From the statements made by the participants and from the observations made by the research team, we conclude that the sociocultural characteristics of the patients are very relevant to the communication of bad news, both in how it is received (non-verbal behaviour) and in the coping strategies then adopted.
### Reactions to bad news {#Sec20}
The healthcare professionals in our study population do not find it difficult to identify the emotions aroused in patients on receiving bad news, because certain patterns tend to appear repeatedly. Moreover, the professionals consider it positive to encourage the expression of these emotions.
### Demand for information {#Sec21}
As with emotional responses, the first questions asked by women in this situation are very familiar to healthcare workers (*"Why? What did I do wrong? Now what? Does this happen often?").* When an adverse prenatal diagnosis is made, patients are advised to request a second medical opinion before making a final decision regarding their pregnancy, and are discouraged from seeking information on the internet, because of its unreliability. Patients are warned that the only dependable information is that provided by the health team.
### Influence of professionals on decision-making {#Sec22}
The professionals realise that their opinions could affect patients' decisions about their pregnancies, depending on the communicative style employed. The way in which questions such as the prognosis and possible alternatives are addressed could facilitate or hinder a satisfactory resolution of the situation. Variability among the professionals involved in this healthcare can also make communication difficult.
### Psychosocial support {#Sec23}
Healthcare professionals are well disposed to accompany these patients and offer them support, but believe they lack training to respond adequately to the patients' grieving. The general opinion is that handling difficult situations requires the intervention and support of a psychologist specialised in providing this sort of assistance.
### Influence of technology {#Sec24}
The professionals also referred to being stressed by the bureaucratic and treatment overload they are obliged to accept in modern treatment contexts, together with the ever-greater dependence on technology. According to the professionals, these factors prevent them from dedicating sufficient time to address human concerns and to prevent the medical act from becoming increasingly impersonal (Table [4](#Tab4){ref-type="table"}).Table 4Category.- Patient - Healthcare professional interactionSubcategory: profile of the patients\
Verbatim***OBST S13*** *: "Young women, between 16 and 44 years old, generally healthy for pregnancy, childbirth and postnatal care. Regarding socio-cultural level, there are all types, from low socio-cultural level to middle and high levels, immigrants, Spanish natives, Asian, European, African \... a multicultural population."****OBST S8*** *: "Very heterogeneous due to the variety of races."****OBST S7*** *: "Because what I say may lead to the pregnancy being interrupted, it's necessary to know that not all cultures conceive or face this prospect in the same way."*Subcategory: reactions to bad news\
Verbatim***MID S3*** *: "At first there is a state of shock, a sense of unreality."****OBST S3*** *: "They respond with pain, crying, anguish, suffering, and the feeling of enormous disappointment."****OBST S8*** *: "Although the foetus referred to in the bad news is the fruit of two people, the father and the mother, the mother's response is usually much more emotional, and the role of the father automatically becomes that of consoling the mother."*Subcategory: demand for information\
Verbatim***NASST S6****: "Why*? *What have I done? Is it common? Finding out the cause and trying to determine if they are responsible. Is it something I took? I made an effort (\...)"****NASST S5****: "(\...) Now what*? *What can be done? What do you suggest? What would you do if it happened to you?"****NUR S1*** *: "I think a second opinion would be a good idea, but it is very important to know who to ask."****OBST S4*** *: "They resort to the internet, but they don't know how to filter the information and what information has been scientifically proven."****MID S3*** *: "In the private sector, they think they are better looked after because they are given more time and attention. Perhaps we should improve things in this area."*Subcategory: influence of professionals on decision-making\
Verbatim***OBST S14*** *: "(\...) you are often recommend what they should do. So, the way we give the news can make their decision go one way or the other."****NUR S2*** *: "The lack of social skills hinders the active participation of these patients in decision making."****OBST S13*** *: "I try to be as aseptic as possible, to respect their autonomy, giving them the consent form to sign ..."*Subcategory: psychosocial support\
Verbatim***OBST S10*** *: "(\...) some patients do need it, because they collapse, they go home and for months they feel very bad and don't know who to turn to."****MID S2*** *: "The psychologist should be part of the team, both to help the women and to guide staff, because burnout does happen."****NUR S3*** *: "We don't know what techniques we can use to deal with conflicts that may arise during the clinical relationship."*Subcategory: influence of technology\
Verbatim***OBST S13*** *: "We spend more time using technology than we do listening, looking into people's eyes (\...)"****NUR S3*** *: "An excess of technology can dehumanise the attention we provide."OBST* Obstetrician, *MID* Midwife, *NUR* Nurse, *NASST* Nursing assistant
Discussion {#Sec25}
==========
As is apparent from our findings, although in health care it is frequently necessary to break bad news to patients, this obligation poses a major challenge to doctors and nurses, and can create difficult, painful situations \[[@CR22], [@CR23]\]. The way in which bad news is transmitted affects the patient's understanding of the information received and hence the decisions taken in this respect, the psychological adaptation to new circumstances, participation in the process and any future changes made \[[@CR24]--[@CR26]\].
Difficulties may arise from communicators' insecurity and anxiety, possibly due to inadequate training in communication techniques and care relations. Moreover, healthcare staff may lack the knowledge and skills needed to assess patients' information needs and to motivate their active participation in decision making \[[@CR15], [@CR27]--[@CR29]\].
In this respect, intuition and/or experience are not sufficient in themselves. Communication is not a gift but a skill that can be learned \[[@CR30]\] and for which training must be provided, because it does not necessarily improve with experience \[[@CR31]\]. The doctors, nurses and nursing assistants who took part in this study all agree that in developing their professional competence, they learned to communicate with patients by means of trial and error and by imitation, from observing the actions of colleagues with more experience. None of the medical workers taking part in our study had received a refresher course or specific training in this respect, a shortcoming that has also been reported in previous research \[[@CR32]--[@CR35]\].
Communication skills should be included as part of the training of healthcare personnel, together with the clinical competence specific to each branch of the profession \[[@CR36]\]. Such training can enhance empathy in carers, helping them evaluate patients' expectations, offer appropriate support, reduce emotional distress and foster compliance with clinical guidelines \[[@CR26]\]. Indeed, good communication is an ethical and legal imperative \[[@CR34], [@CR37], [@CR38]\].
In line with previous studies \[[@CR33], [@CR34], [@CR39]\], we believe that other important aspects to be addressed include non-verbal language and the environment in which the bad news is to be communicated: this should be comfortable and quiet and enable privacy. However, this is not always the case, and organisational and structural problems have been identified. Consultation areas often fail to provide the above-mentioned characteristics, and noise, interruptions and lack of time due to the care burden all impede the creation of the therapeutic rapport between medical worker and patient that is necessary for successful collaboration between them.
Midwives, nurses and nursing assistants receive training in providing specific support to maternity patients, reinforcing personal qualities traditionally associated with their profession, such as closeness, kindness and sympathy. Corroborating the findings of earlier research, our results show that healthcare personnel are equipped with the necessary skills to produce a good nurse-patient relationship, although it has also been found that these capabilities are often lost if they are not refreshed during later professional practice \[[@CR40]\].
While nurses must often play the role of communicators of bad news, many obstetricians consider this to be a function that is outside their sphere of competence. Nursing staff and midwives, however, are usually committed to working as a team, and the multidisciplinary approach has been shown to be the most effective means of communicating bad news. Indeed, if health teams do not function in a well-integrated way, the patient may receive differing or even contradictory information \[[@CR33], [@CR41]\].
To overcome barriers to communication, healthcare personnel should develop the ability to express empathy, closeness and solidarity with patients' emotions, and also possess active listening skills -- in areas such as paying careful attention and manifesting availability to help -- together with self-assuredness, transmitting a sense of security on the basis of well-grounded opinions \[[@CR3], [@CR39], [@CR42], [@CR43]\].
The lack of training in the above skills to foster effective communication is often aggravated by an absence of feedback and by insufficient time to offer more support. As a result, the relationship with the patient is very limited despite the wish to provide high quality care. A possible problem in the carer-patient relation is the tendency of carers, in some cases, to (mis)interpret the wishes and needs of patients; in consequence, the response made may not meet the patient's expectations, but correspond to what the carer believes appropriate. A good communicator should clarify matters with the patient, providing feedback to ensure that what is clear to one party is equally clear to the other \[[@CR5], [@CR6]\].
The acquisition of communication skills is hampered when there is no relationship between the learning process and the context in which the work must be carried out, when there is little or no flexibility in the scheduling of courses or workshops, when there is insufficient institutional recognition and when there are few opportunities for training to take place. Problems may also arise from a fear of becoming too involved when dealing with personal aspects such as feelings and emotions \[[@CR44]\].
Health professionals may adopt different attitudes if they lack the skills to handle the emotional responses of patients who have received bad news. On the one hand, some carers believe that their responsibility is limited to addressing physical problems; in consequence, they will emphasise the development of skills related to the use of instruments developed in biomedicine, in order to provide a faster and more accurate diagnosis \[[@CR1]\]. In monitoring the progress of a pregnancy control, such a carer's entire attention could be focused on confirming or ruling out the presence of biological problems in the foetus. Accordingly, ultrasound examinations, screening tests, the signing of consent forms and entering medical records into computer files would occupy most of the time allocated to the consultation, leaving hardly any opportunity for a face-to-face exchange of views \[[@CR10]\]. A carer may be highly skilled in the application of certain techniques, but this will be of little use if effective communication cannot be established \[[@CR31]\]. In fact, technology should be auxiliary to the carer's daily work \[[@CR45]\]. Patient-centred care means understanding the human being from a biopsychosocial approach, recognising the need to consider not only the illness but also the patient's personal experience, and to understand her as a person with emotions and private concerns (the psychological sphere) in multiple aspects of life, including work, family, and the partner (social sphere) \[[@CR25], [@CR26]\].
On the other hand, time pressures and limitations are sometimes cited to justify the lack of attention paid to psychosocial aspects of health care \[[@CR10]\]. In contrast to this attitude among healthcare personnel, evidence suggests that patients strongly believe there is a need to investigate their unexpressed concerns, to teach them how to evaluate the information provided and to adopt appropriate measures, on the basis of personalised recommendations \[[@CR46]\]. Our review of the literature and our analysis of the results obtained lead us to conclude that in a context in which patients can express their concerns and fears, motivated by the health carer's open, sympathetic attitude, therapeutic communication can be established and the necessary emotional support supplied \[[@CR47]\]. However, doctors and nurses do not usually consider the patient's mood before communicating a diagnosis, or inquire about recent stressful episodes (such as the death of a loved one, or the presence of a severe illness in the patient or in a close relative), despite research evidence that the accumulation of stress-provoking experiences shortly before a traumatic event can increase the incidence of post-traumatic stress disorders \[[@CR48]\].
Difficulties in communication may also be due to an unexpected diagnosis, without previous indications. When a diagnosis of this type must be confirmed, the health carer often experiences anxiety, a feeling of responsibility and a fear of censure, while being pressured to supply a rapid, convincing explanation and at the same time respond to the patient's emotional reaction \[[@CR8]\].
At other times, difficulty in transmitting an adverse diagnosis or a poor prognosis is the result of a heavy workload, together with pressing demands by patients and their families for information \[[@CR49]\]. Studies have shown that when information is transmitted to patients and their families in sufficient quantity and quality, their anxiety is reduced and, in general, better and faster recovery is achieved and patient/carer collaboration is enhanced \[[@CR23], [@CR32]\].
Feelings of frustration and helplessness when the carer is unable to prevent, halt or reverse a negative outcome also hamper communication, especially when the therapeutic options are limited or non-existent. The carer must then seek to provide comfort in a situation that does not offer grounds for being hopeful \[[@CR26], [@CR50]\]. Feelings of frustration and powerlessness may be compounded by legal concerns, with the worry that an unsatisfied patient may present an official complaint. The judicialisation of healthcare issues may generate the notion that every human being has the right to be healed and that any failure in this respect must be due to an error, which should be punished. In the health service, more complaints are made regarding the quality of information received than any other aspect of health care \[[@CR10]\]. Moreover, the provision of informed consent does not always guarantee the reality of bidirectional communication, but may serve only as a legal safeguard \[[@CR51]\].
Our results also show that some healthcare professionals, due to the experience of patients' suffering and to taboos regarding death, erect barriers to communication through automated responses and patterns of avoidance, especially when there is the possibility of transferring responsibility to other carers \[[@CR52]\]. This type of behaviour may arise due to a lack of support system for the healthcare team, or to the absence or outdated status of action protocols \[[@CR50]\].
Language and cultural differences are the most common types of communication barrier, due to the multilingual and multicultural nature of the treatment population \[[@CR53]\]. The two most important causes of ineffective communication are the precarity of information provided to patients and the absence of comprehension \[[@CR54], [@CR55]\].
The link between the principal investigator and the institution in which the study was performed may be regarded as a limitation of this study, insofar as it may have influenced the interpretation of certain professional practices. On the other hand, this association facilitated access to a wide range of scenarios in which patients and carers interact, thus providing an authentic outlook on healthcare practice.
Another potential issue is that, given the inherent properties of the qualitative method applied and the local nature of this study, the findings obtained might not be readily extrapolated to other contexts (although the method used can be transferred without difficulty). The effect of subjectivity means that a given phenomenon may be perceived, interpreted and experienced differently from one individual to another. Therefore, the interpretations of our participants, regarding the specific problem addressed in this study, will not necessarily be shared by other professionals, interacting with patients in different contexts.
Conclusions {#Sec26}
===========
The increasing dependence of certain diagnostic procedures on technological resources may be detrimental to interpersonal relationships, making them cold and distant. For healthcare personnel, the human quality of their profession has deteriorated, mainly due to the heavy caseloads experienced, which increasingly limit the time that can be spent with each patient. In order to improve communication, more attention should be paid to the human and spiritual dimensions of healthcare, giving greater weight to empathy, authenticity and listening (without imposing one's own interpretation). The analysis performed leads us to draw the following conclusions: a different model of clinical relationship should be promoted, based on shared decision making, and greater clarity should be granted to the functions of the multidisciplinary team with respect to the patient's grieving when a pregnancy is interrupted. To achieve these goals, protocols should be implemented to ensure comprehensive care provision, addressing not only the biological sphere but also psychosocial concerns. In this respect, too, specific training should be provided, at undergraduate and postgraduate levels, in social skills and cultural competence. In short, this study identifies possible areas of improvement related to the interventions of healthcare personnel and to the organisation of the institution itself, with particular respect to the communication to patients of an adverse prenatal diagnosis.
The superscripts correspond to the authors who have made this part of the methodology.
We thank all the staff of the Costa del Sol Health Agency who agreed to participate in this study. We also thank the research team of the Hospital of the Costa del Sol for their support. In addition, we are grateful for the contribution made to this research project by the University of Malaga.
Funding {#FPar1}
=======
This study was partially supported by the Costa del Sol Health Care Agency, which provided funds for the dissemination of the study results.
Availability of data and materials {#FPar2}
==================================
The datasets generated and analysed in the present study are not publicly available, in accordance with Spanish legislation on personal data protection (Act 15/1999), but can be obtained from the corresponding author on reasonable request.
JAC proposed the concept for the study, conducted and transcribed the interviews, and developed the first draft of the manuscript. MLA, MPR and IMMG contributed ideas and discussed the results of the study. All the co-authors read, critically commented on and revised the successive drafts of the manuscript. They also approved the final version before submission to the journal.
Ethics approval and consent to participate {#FPar3}
==========================================
This study was approved by the Local Research Ethics Committee Costa del Sol (Spain) CEI (003_mayo_PR2-Malas Noticias) on 01 June 2015, and was carried out in accordance with the ethical principles as set out in the Declaration of Helsinki. Written consent was obtained from all the participants in the study.
Consent for publication {#FPar4}
=======================
Not applicable in this section.
Competing interests {#FPar5}
===================
The authors declare that they have no competing interests.
Publisher's note {#FPar6}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
Editorial Note on the Review Process
====================================
[F1000 Faculty Reviews](http://f1000research.com/browse/faculty-reviews) are commissioned from members of the prestigious [F1000 Faculty](http://f1000.com/prime/thefaculty) and are edited as a service to readers. In order to make these reviews as comprehensive and accessible as possible, the referees provide input before publication and only the final, revised version is published. The referees who approved the final version are listed with their names and affiliations but without their reports on earlier versions (any comments will already have been addressed in the published version).
The referees who approved this article are:
1. Aliya A. Khan, Medicine, Divisions of Endocrinology and Metabolism and Geriatric Medicine, McMaster University, Hamilton, Canada [^2]
2. Maria Luisa Brandi, Department of Surgery and Translational Medicine, Bone Metabolic Diseases Unit, University of Florence, Florence, Italy [^3]
3. Bart Clarke, Division of Endocrinology, Diabetes, Metabolism, and Nutrition, Mayo Clinic College of Medicine, Rochester, MN, USA [^4]
Introduction
============
Hypoparathyroidism is a rare endocrine disorder in which the production of parathyroid hormone (PTH) by the parathyroid gland is absent or inappropriately low ^[@ref-1],\ [@ref-2]^. The prevalence of hypoparathyroidism in the United States and Europe is estimated at 23--37 per 100,000 individuals ^[@ref-3],\ [@ref-4]^. The most common cause of hypoparathyroidism is damage to or removal of the parathyroid glands during neck surgery, accounting for approximately 75% of cases ^[@ref-2],\ [@ref-4]^. Autoimmunity targeting the parathyroid gland and, rarely, genetic disorders can also lead to hypoparathyroidism ^[@ref-1],\ [@ref-5]^. DiGeorge syndrome, the most common genetic cause, is associated with a microdeletion in chromosome 22q11.2 and parathyroid hypoplasia ^[@ref-6],\ [@ref-7]^. Other genetic causes include autosomal dominant hypocalcemia (ADH), in which an activating mutation of the calcium-sensing receptor (CaSR) decreases its set-point ^[@ref-8]^, and autoimmune polyendocrine syndrome type 1, in which a mutation in the autoimmune regulator gene ( *AIRE*) leads to destruction of the parathyroids and other endocrine glands ^[@ref-9]^.
PTH plays a fundamental role in mineral homeostasis by promoting renal reabsorption of calcium and stimulating renal phosphate excretion. It also promotes conversion of 25-hydroxyvitamin D to 1,25-dihydroxyvitamin D (1,25\[OH\] ~2~D), the active form of vitamin D, which increases the absorption of calcium and phosphate from the gastrointestinal tract. Moreover, PTH is a powerful regulator of bone turnover, such that PTH deficiency leads to decreased bone turnover. Thus, patients with hypoparathyroidism experience hypocalcemia, hyperphosphatemia, hypercalciuria, reduced levels of 1,25\[OH\] ~2~D, and abnormally low bone turnover, resulting in overly mineralized bone ^[@ref-1],\ [@ref-2],\ [@ref-10],\ [@ref-11]^.
The diagnosis of hypoparathyroidism requires confirmed hypocalcemia in the presence of undetectable or inappropriately low levels of endogenous PTH and the absence of hypomagnesemia, the latter condition being a reversible cause of hypoparathyroidism; other potentially reversible causes include iron or copper overload and autoimmune causes ^[@ref-1]^. Hypocalcemia typically manifests as perioral numbness, paresthesias, neurocognitive deficits, weakness, and carpopedal muscle spasms, although potentially life-threatening complications, such as cardiac arrhythmias, laryngeal spasm, tetany, and seizures, can also occur ^[@ref-1],\ [@ref-2],\ [@ref-12],\ [@ref-13]^.
Conventional therapy
====================
Conventional management of hypoparathyroidism consists of oral calcium (e.g. calcium carbonate or citrate) and active vitamin D treatments (e.g. calcitriol), as well as thiazide diuretics to increase calcium reabsorption at the distal tubule, and magnesium supplementation as needed. The goal is to maintain serum calcium levels just below or within the lower normal range. Additional goals include preserving a normal level of serum phosphate, preventing an elevation of the serum calcium-phosphorus product, minimizing hypercalciuria, and avoiding kidney stones and mineralization of soft tissues.
Although conventional therapy increases intestinal calcium absorption and corrects hypocalcemia, it does not replace other functions of PTH and, in the absence of PTH's calcium-retaining effect in the kidney, can lead to hypercalciuria ^[@ref-1],\ [@ref-2],\ [@ref-10],\ [@ref-14]--\ [@ref-16]^. Other limitations of conventional therapy include unpredictable occurrences of hypocalcemia and hypercalcemia, increased calcium-phosphorus product, and long-term complications such as ectopic calcifications, nephrocalcinosis, nephrolithiasis, and renal dysfunction ^[@ref-1],\ [@ref-2],\ [@ref-10],\ [@ref-14]--\ [@ref-16]^. Symptoms of hypocalcemia and reduced quality of life (QoL) are common among patients managed with conventional treatment ^[@ref-17]--\ [@ref-20]^.
PTH therapy
===========
Treatment of hypoparathyroidism with intermittent PTH injections has been an important advance. Studies by Winer *et al*. with PTH(1-34) showed that subcutaneous injection could maintain normocalcemia in hypoparathyroid adults ^[@ref-21]^ and children ^[@ref-22]^ as well as calcitriol for as long as 3 years ^[@ref-23],\ [@ref-24]^ and that twice-daily dosing reduced the total amount of PTH required ^[@ref-25]^. Recombinant human PTH \[rhPTH(1-84)\] is full-length PTH that was approved in the United States in 2015 as a once-daily subcutaneous administration as an adjunctive treatment of adults with hypoparathyroidism who could not be well controlled on conventional therapy alone ^[@ref-26]^. Safety and efficacy were demonstrated in placebo-controlled and open-label studies ^[@ref-27]--\ [@ref-30]^. In the pivotal REPLACE study, a 24-week, double-blind, placebo-controlled, randomized phase III study conducted with 134 patients, 53% of patients receiving rhPTH(1-84) versus 2% of patients receiving placebo met the primary study end point (≥50% reduction in calcium and calcitriol doses with maintenance of normal serum calcium) ^[@ref-28]^.
Renal complications
===================
Without the renal calcium-retaining effects of PTH, conventional therapy often leads to elevated urinary calcium excretion and long-term renal complications, including nephrocalcinosis, nephrolithiasis, and, consequently, the development of chronic kidney disease ^[@ref-15],\ [@ref-16]^. In a case-control study, using the Danish National Patient Registry, Underbjerg and colleagues found decreased eGFR (\<60 mL/minute) in 21% of hypoparathyroid (n = 431, mostly postsurgical) patients ^[@ref-31]^. This increased risk of renal disease was significantly associated with a higher number of hypercalcemic episodes and a higher calcium-phosphate product ^[@ref-31]^. An even higher rate of kidney disease was shown in a retrospective chart review of 120 mostly postsurgical hypoparathyroid patients at a tertiary care center ^[@ref-15]^. This report showed two- to 17-fold greater rates of developing chronic kidney disease (stage 3--5) than age-appropriate adjusted rates, with 41% of hypoparathyroid patients having chronic kidney disease ^[@ref-15]^. Specific risk factors for a lower eGFR included age, duration of disease, and proportion of time with relative hypercalcemia ^[@ref-15]^. Nonsurgical hypoparathyroid patients were also shown to experience renal disease, with 14% out of 165 patients having an eGFR \<60 mL/minute; a younger age at presentation and a higher serum calcium-phosphorus product increased the risk for nephrocalcinosis ^[@ref-32]^. Taken together, these studies show that renal disease is a frequent consequence of hypoparathyroidism and is likely mediated by exposure to hypercalcemia and abnormal calcium-phosphate homeostasis. It is unknown whether reducing serum phosphate levels, either by diet or phosphate binders, minimizes renal complications.
PTH treatment, by increasing renal calcium reabsorption, would be expected to decrease urinary calcium excretion in hypoparathyroidism. Disappointingly, this effect has not been demonstrated in a RCT. In the REPLACE study, in both rhPTH(1-84) and placebo groups, urinary calcium excretion declined from baseline, but, at week 24 in both groups, the mean urinary calcium excretion remained above the upper limit of normal for women and near the upper limit of normal for men ^[@ref-28]^. However, the rhPTH(1-84) group had significant reductions in serum phosphate and the serum calcium-phosphate product throughout the study compared with patients receiving placebo ^[@ref-33]^.
In the 60-month open-label extension of REPLACE (n = 40), the mean urinary calcium level declined into the normal range for men and women ^[@ref-30]^. The serum phosphate and calcium-phosphate product also declined, while serum calcium, creatinine, and eGFR remained unchanged ^[@ref-30]^. Similar beneficial effects on urinary calcium excretion were seen in the other long-term open-label study of rhPTH(1-84), in which urinary calcium excretion declined significantly over 8 years ^[@ref-27]^. In that study, serum phosphate and the calcium-phosphate product did not decrease but remained within the normal range, as did eGFR. Overall, these long-term uncontrolled studies suggest that urinary calcium excretion and serum phosphate levels remain decreased with prolonged rhPTH(1-84) therapy. Intuitively, reduced urinary calcium excretion would be expected to decrease the risk of renal dysfunction in hypoparathyroid patients; rigorously controlled clinical trials are required to validate this hypothesis.
Quality of life complications
=============================
Hypoparathyroid patients typically suffer from symptoms related to increased neuromuscular excitability, such as parasthesias and carpopedal spasm. In addition, hypoparathyroid patients often describe neurocognitive complaints, including anxiety, depression, fatigue, and various cognitive deficits ^[@ref-13],\ [@ref-18],\ [@ref-34],\ [@ref-35]^. They have lower QoL scores, using validated tools, mostly the Short Form Health Survey 36 (SF-36), a self-administered 36-item questionnaire, compared with reference norms ^[@ref-18],\ [@ref-19],\ [@ref-34],\ [@ref-36],\ [@ref-37]^. Both mental and physical domains are affected, although the mechanisms for the deficits are uncertain. It may be that the physical complaints are attributable to reduced muscle function ^[@ref-34]^, which might in turn be associated with fluctuations in serum calcium. The mental QoL complaints might be explained by effects of calcium on the central nervous system, as supported by findings in primary hyperparathyroidism of neuropsychological symptoms ^[@ref-38]^. Interestingly, QoL seems to be more reduced in patients with postsurgical hypoparathyroidism compared with non-surgical hypoparathyroidism ^[@ref-39]^. This might be explained by postsurgical patients also having hypothyroidism ^[@ref-39]^. Alternatively, patients with life-long nonsurgical hypoparathyroidism might be less sensitive to changes in their QoL, having never experienced a full sense of well-being.
The effects of PTH treatment on QoL are unclear. Anecdotally, patients report marked improvements, but a significant favorable effect of PTH-replacement therapy was not shown in double-blind randomized trials compared with controls receiving conventional therapy ^[@ref-20],\ [@ref-36]^. In the REPLACE study, SF-36 scores did not differ between the two treatment arms, rhPTH(1-84) versus placebo, at 24 weeks ^[@ref-36]^. One can speculate that this negative finding was due to the lack of sustained PTH elevations with the frequency of PTH administration used. Specifically, once-a-day rhPTH(1-84) injections can cause fluctuations over the 24-hour period ^[@ref-40]^, with an initial rise leading to transient hypercalcemia and possible symptoms of nausea, poor concentration, and polyuria, followed by a fall in serum calcium levels with possible hypocalcemia symptoms.
Open-label uncontrolled data regarding PTH treatment and QoL are more positive ^[@ref-41]^. In a long-term open-label study of 28 hypoparathyroid patients treated with rhPTH(1-84) therapy over 8 years, improvement using the SF-36 was observed, particularly in subjects with impaired SF-36 scores at baseline and in those whose requirements for conventional therapy decreased substantially ^[@ref-19]^. It is possible that PTH itself, rather than restoration of eucalcemia, may improve QoL. Support for this idea includes the observation that PTH can cross the blood--brain barrier ^[@ref-42]^ and can stimulate PTH2 receptors that are in the brain ^[@ref-43]^. These PTH2 central nervous system receptors have been associated with areas involved in the regulation of depression and anxiety in animals ^[@ref-44],\ [@ref-45]^.
Importantly, a disease-specific questionnaire is not available for hypoparathyroidism. Most studies of PTH effects on QoL used the SF-36, which is subjective and does not adequately measure cognitive deficits, fatigue, physical endurance, or muscle strength. A single instrument that can characterize and measure the burden of disease experienced by patients with hypoparathyroidism is needed. A subjective "Hypoparathyroidism Symptom Diary" is being developed and tested ^[@ref-46]^. This is a 13-item instrument, which assesses symptoms including muscle cramping, tingling and muscle spasms/twitching, fatigue, and cognition, as well as anxiety, sadness, and depression, and impacts on sleep, ability to exercise, ability to work, and family relationships ^[@ref-46]^. The development of a validated tool to assess disease burden and efficacy of treatment in hypoparathyroidism will be a key advance.
Skeletal complications
======================
Hypoparathyroidism is associated with low bone turnover (the process of coupled bone formation and bone resorption), which is associated with increased bone mineral density and abnormal bone microarchitecture ^[@ref-47],\ [@ref-48]^. The reduction in bone formation is demonstrated by the decrease of tetracycline labelling in bone biopsies compared with controls ^[@ref-49]^. Notably, fracture risk in hypoparathyroidism is uncertain. Case-control studies demonstrate no differences in overall hypoparathyroidism fracture rates ^[@ref-16],\ [@ref-35]^, while other reports show an increased fracture risk at the spine and upper extremities ^[@ref-13],\ [@ref-15],\ [@ref-50]^.
With rhPTH(1-84) therapy, open-label data show an initial large increase in bone turnover markers at 1 year of therapy, followed by a new steady state that remains above baseline with continued therapy ^[@ref-30],\ [@ref-51]^. In an 8-year study of rhPTH(1-84), lumbar spine bone mineral density increased in the early years and plateaued after 4 years of rhPTH(1-84) while an increase in total hip bone mineral density was seen in the later years of PTH therapy. Femoral neck bone mineral density remained unchanged throughout the study, while one-third radial bone mineral density declined modestly ^[@ref-27]^. The increase at the trabecular-enriched lumbar spine and the decrease at the cortical 1/3 radius were reminiscent of the patterns observed with intermittent PTH treatment for osteoporosis. Histomorphometric analysis of bone biopsies showed that intra-trabecular tunneling was present, with increases in cancellous bone volume and trabecular number and cortical porosity ^[@ref-52]^, suggesting exuberant stimulation of bone remodeling. Fracture data are needed to determine whether bone fragility is present in the context of long-term rhPTH(1-84) use.
Continuous PTH administration
=============================
As noted above, daily administration of rhPTH(1‐84) did not demonstrate a reduction in hypercalciuria compared with conventional therapy in the REPLACE registration RCT ^[@ref-28]^. This shortcoming might have been due to the relatively short half-life of rhPTH(1‐84) with daily subcutaneous administration. PTH levels are transiently high after injection but then drift downwards during the hours preceding the subsequent injection ^[@ref-53]^. Without continuous PTH exposure to stimulate renal calcium reabsorption, urinary calcium excretion increases.
The benefits of continuous PTH exposure were shown in pump studies by Winer and colleagues. Continuous subcutaneous infusion of PTH(1‐34) with an insulin pump to hypoparathyroid patients mimicked endogenous PTH secretion more closely than intermittent PTH injections ^[@ref-54],\ [@ref-55]^. In both hypoparathyroid adults and children, the pump delivery method led to normalization of serum calcium, urinary calcium excretion, and bone remodeling as measured by bone turnover markers. Moreover, serum phosphate levels normalized in adult hypoparathyroid patients after continuous subcutaneous infusion of PTH(1‐34), although not in hypoparathyroid children. Magnesium levels improved significantly in adults and children. Notably, continuous infusion of PTH(1‐34) was superior by all measures compared with a twice-daily subcutaneous dosing regimen with PTH(1-34), despite a \>60% lower daily dose of PTH with infusion ^[@ref-54],\ [@ref-55]^. These studies suggest that continuous exposure to PTH provides a highly physiologic effect to reverse the biochemical derangements that are a hallmark of hypoparathyroidism.
An important gap in the management of hypoparathyroid patients is the lack of a technology to continuously monitor endogenous calcium levels. Given that calcium fluctuations can be triggered by exercise, general illness, or unpredictable causes ^[@ref-1]^, hypoparathyroid patients would benefit from knowing their calcemic variability in real time, giving them a chance to adjust their treatment individually and dynamically. A device for monitoring calcium to allow for dose adjustments in response to fluctuations in blood calcium would facilitate such titration. Such a continuous calcium sensor in combination with PTH delivery by a pump device could constitute a transformative closed-loop "artificial parathyroid" system for hypoparathyroid patients.
Current and future therapies
============================
rhPTH(1-84), or Natpara, was recalled in the United States in September 2019 owing to a potential problem: over the course of treatment, the rubber septum of the Natpara cartridge is punctured by a needle to obtain the daily dosage, and when this occurs repeatedly, there is a possibility that small rubber particles may detach from the septum into the cartridge ^[@ref-56]^. A "Special Use Program" is available to patients previously prescribed Natpara who are at extreme risk of life-threatening complications as a result of discontinuation of treatment. This program requires that product usage be limited to a single dose per cartridge, instead of 14 doses per cartridge, to minimize the potential of particle formation caused by repeat punctures ^[@ref-56]^. Importantly, PTH cessation can lead to rebound hypocalcemia, likely due to the reversal of the enlarged remodeling space favoring bone formation as the skeleton returns to a low turnover state ^[@ref-57]^.
Off-label use of teriparatide (rhPTH\[1-34\]) has been used in hypoparathyroid patients. Although it is not FDA-approved for the treatment of hypoparathyroidism, there is experience with its use in clinical trials ^[@ref-21]--\ [@ref-25]^. Individualized dosing, administered as twice-daily or even thrice-daily injections, might be necessary because of the relatively short effect of teriparatide to raise serum calcium levels ^[@ref-58]^.
A potential new therapy is TransCon PTH, an inactive prodrug of PTH(1-34) that is designed to achieve an infusion-like profile by liberating active PTH in a sustained fashion ^[@ref-59]^. PTH(1-34) is transiently bound via a linker to a chemically inert carrier that shields PTH from binding to the PTH type 1 receptor (PTHR1) and prolongs the peptide's circulation half-life due to reduced renal clearance ^[@ref-59]^. In euparathyroid and hypoparathyroid rats and in euparathyroid monkeys, it led to steady concentrations of PTH within the normal range, with persistent increases in serum calcium and decreases in serum phosphate levels ^[@ref-59]^. The results of a phase I trial with TransCon PTH in healthy volunteers demonstrated a markedly prolonged half-life of free PTH of \~60 hours ^[@ref-60]^. Notably, a net increase in bone resorption with TransCon PTH administration has been observed ^[@ref-59]^. Although this could theoretically lead to a primary hyperparathyroidism-like profile, increased resorption could have a salutary effect to decrease the high bone mineral density and increase the low bone turnover rate of hypoparathyroidism. A phase II trial with TransCon PTH is on-going.
An additional potential therapy is PCO371, a novel orally active compound that acts as an agonist of PTHR1 ^[@ref-61]^. In a series of *in vitro* and *in vivo* experiments, Tamura *et al*. demonstrated that oral PCO371 has similar biochemical and renal effects as PTH injections, particularly in hypoparathyroid rats, restoring serum calcium levels and lowering serum phosphate levels without increasing urinary calcium excretion ^[@ref-61]^. Although it was less potent than human PTH (hPTH)(1-34) *in vitro*, the *in vivo* effects were more long-lasting than those of hPTH(1-34) and hPTH(1-84), consistent with greater bioavailability. PCO371 is currently undergoing clinical testing. If effective, this oral PTH analogue could overcome concerns about injectable use and be a future treatment option for hypoparathyroid patients.
A further creative approach may be a long-acting analogue of PTH (LA-PTH). This modified peptide mediates prolonged and enhanced effects to increase blood calcium levels in hypoparathyroid rats as compared to PTH(1-34) and PTH(1-84) ^[@ref-62],\ [@ref-63]^. The prolonged signaling action appears to involve a mechanism of pseudo-irreversible binding to a distinct PTHR1 conformation (R ^0^), which maintains high affinity for the ligand through multiple rounds of G protein coupling ^[@ref-64]^. The biological actions of LA-PTH in humans remain to be investigated.
Conclusion
==========
Our knowledge of hypoparathyroidism has increased over the past few decades, yet unresolved questions remain. It is unknown whether the long-term complications of hypoparathyroidism can be prevented or reversed. Moreover, key tools are lacking, such as a validated disease-specific tool to measure QoL and a continuous calcium sensor to enable real-time dose titration. PTH therapy is likely to be most effective when administered in a physiologic manner that minimizes serum and urinary calcium fluctuations, yet large and long-term studies testing continuous PTH exposure are lacking. Answers to these and other key questions will ultimately improve our understanding and management of this rare disease.
[^1]: **Competing interests:**Mishaela Rubin receives research funding from Takeda and Ascendis.
[^2]: No competing interests were disclosed.
[^3]: No competing interests were disclosed.
[^4]: No competing interests were disclosed.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Artemisinin (ART) and related compounds provide the main class of anti-malarial drugs, and ART resistance in *Plasmodium falciparum* is one of the greatest threats to global efforts to control, eliminate and eradicate malaria. To forestall emergence and spread of ART resistance it was recommended that ART and its derivatives be used only in combination with a partner drug as an ART combination therapy (ACT), with over 400 million ACT treatments dispensed annually. It is conservatively estimated that 116,000 additional deaths would occur annually in the event of widespread ART resistance, with annual health costs of US\$32 million and productivity losses exceeding US\$385 million \[[@CR1]\]. The magnitude of the health and economic threat posed by ART resistance serves as an urgent call to action to develop strategies that circumvent its spread \[[@CR2]\]. Doing so requires understanding the underlying mechanisms of ART resistance.
Currently, ART resistance is prevalent across the Greater Mekong Sub-region (GMS) and centered on Cambodia, where it was first detected in 2007 \[[@CR3], [@CR4]\]. Concern for the spread of ART resistance outside of Southeast Asia led to the Tracking Resistance to Artemisinin Collaboration (TRAC) study that assessed and tracked ART resistance across 15 sites both in Asia and Africa \[[@CR5]\]. Using parasite clearance as a measure of resistance, the TRAC project confirmed that ART-resistant *P. falciparum* was established in Cambodia, Laos, Myanmar, Thailand, and Vietnam.
Artemisinin resistance is indicated by either delays in parasite clearance from patients, or by increased in vitro parasite survival under dihydroartemisinin (DHA) in ring-stage survival assay (RSA~0--3h~) \[[@CR6]\]. Sequencing parasites selected under increasing ART pressure identified *pfkelch13* as a critical gene for conferring ART resistance \[[@CR7]\]. Multiple mutations have been identified in this gene, and several key single nucleotide polymorphisms (SNPs) have been validated using a gene-editing approach \[[@CR8], [@CR9]\]. However, the function of *Pf*Kelch13 and its role in ART resistance remain unclear.
Recent reports confirm that specific *pfkelch13* mutations confer ring-stage survival \[[@CR7], [@CR8]\], and indicate possible mechanisms of ART resistance, including perturbation of haemoglobin processing \[[@CR10]--[@CR12]\], protein ubiquitination \[[@CR13]\], increased expression of oxidative stress response \[[@CR13], [@CR14]\], unfolded protein response pathways \[[@CR14]\], or phosphatidylinositide 3-kinase pathways \[[@CR15]\]. While these results may all be relevant, a full understanding of the mechanisms involved in ART action and resistance has yet to emerge \[[@CR16], [@CR17]\].
One of the key partner drugs used with ART for malaria treatment in the GMS is piperaquine (PPQ), which has proved well tolerated and highly effective in areas where multi-drug-resistant *P. falciparum* is prevalent. However, emergence of PPQ resistance threatens to undermine this strategy in areas of increasing ART resistance \[[@CR18], [@CR19]\]. Currently PPQ resistance is evident by higher PPQ half maximal effective concentration (EC~50~) values and elevated recrudescence rates in settings where DHA--PPQ is in use and ART resistance is common \[[@CR18], [@CR19]\]. Reported molecular markers of PPQ resistance include SNPs in *pfcrt* and de-amplification of a region on chromosome 5 that includes or is proximal to *pfmdr1* \[[@CR20]\]. PPQ resistance is also associated with an amplification of *plasmepsin II* and *III* \[[@CR21], [@CR22]\]. Yet, to date no mechanism has been established for PPQ resistance. Increasing resistance to both ART and PPQ in the same parasite population has motivated a call for use of triple combination therapy in certain malaria-endemic settings where ACT may have reduced efficacy \[[@CR18], [@CR23]\]. Specifically, mefloquine (MFQ) has been suggested for combination with ART and PPQ given the inverse effects PPQ and MFQ pressure have on *pfmdr1* copy number variation (CNV), as PPQ negatively selects *pfmdr1* CNV, while MFQ positively selects *pfmdr1* CNV \[[@CR18]\].
As part of the TRAC collaboration, 157 cryopreserved parasites were obtained from two sites in western Cambodia with a range of in vivo clearance phenotypes. Of these parasites, 68 were culture-adapted, and a sub-set of 36 parasites evaluated for their RSA~0--3h~ phenotype. Using a high-resolution melt (HRM) genotyping assay for the most common *pfkelch13* mutation (C580Y) and Sanger sequencing, *pfkelch13* propeller mutations were tested for associations with both parasite in vivo clearance half-lives and in vitro parasite RSA~0--3h~ survival rates. The analysis (1) identified parasites lacking *pfkelch13* mutations, but exhibiting increased RSA~0--3h~ survival phenotype (referred to as discordant parasites), suggesting that loci other than *pfkelch13* may be involved in ART resistance; (2) demonstrated a new association between D584V and increased ring-stage survival; and, (3) detected a PPQ-resistant isolate among these parasites, consistent with other reports from this malaria-endemic region. Although there is a strong association between *pfkelch13* mutations and increased ring-stage survival, mutations in *pfkelch13* are not necessary for this ART resistance phenotype. Furthermore, in vitro PPQ resistance is present among these culture-adapted Cambodian parasites, which will enable further investigation of PPQ resistance mechanisms.
Methods {#Sec2}
=======
These parasite samples were obtained with informed consent from patients enrolled in the TRAC study in the Pailin and Pursat sites located in Western Cambodia. Full details of this study, the approvals, and the clinical and laboratory methodologies have been reported in detail elsewhere \[[@CR5]\].
Culture-adaptation and maintenance of TRAC parasites {#Sec3}
----------------------------------------------------
All parasite samples were collected under protocols approved by ethical review boards in Cambodia, at Oxford University and at the Harvard T. H. Chan School of Public Health. Culture-adaptation of parasites was accomplished by thawing cryopreserved material containing infected red blood cells (iRBCs) that had been mixed with glycerolyte. Parasites were maintained in fresh human blood (O+) and Hepes buffered RPMI media containing 12.5% AB+ human serum (heat inactivated and pooled). Cultures were placed in modular incubators and gassed with 1% O~2~/5% CO~2~/balance N~2~ gas and incubated with rotation (50 rpm) in a 37 °C incubator (Additional file [1](#MOESM1){ref-type="media"}).
Sample extraction {#Sec4}
-----------------
Genetic material were extracted from filter papers (Whatman) and culture-adapted material using Promega DNA IQ Casework Pro Kit for Maxwell 16 (Promega Corp., Madison, WI, USA) and Qiagen (QIAmp DNA Blood Mini Kit) commercial kits, respectively, according to manufacturer instructions.
Genotyping {#Sec5}
----------
Development of a high-resolution melt (HRM) assay to screen populations for mutations around amino acid position 580 in the *pfkelch13* locus. The forward primer (5′-GGCACCTTTGAATACCC-3′), reverse primer (5′-CATTAGTTCCACCAATGACA-3′), and unlabeled, blocked probe (5′-AGCTATGTGTATTGCTTTTGAT-block-3′) were amplified asymmetrically at 0.5, 0.1, and 0.4 μM, respectively with 1 ng template DNA. After an initial 2-min hold at 95 °C, 5 or 10 μL reaction mixtures with 2.5 × HRM master mix (BioFire Defense, Salt Lake City, UT. USA) were PCR amplified for 55 cycles: 95 °C for 30 s, 66 °C for 30 s, and 74 °C for 30 s, followed by a pre-melt step of 95 and 28 °C for 30 s each. Products were melted from 45 to 90 °C on a BioFire Defense LightScanner-384 or -32 and analysed using the manufacturer software. Two plasmid controls containing the wild-type and mutant alleles were included as standards for every HRM run. Molecular barcoding was performed as described \[[@CR24], [@CR25]\] to identify monogenomic samples with unique parasite genotypes.
Whole genome sequences {#Sec6}
----------------------
This publication uses sequencing data generated by the Pf3k project \[[@CR26]\]. The variant call files generated from this project were used to identify SNPs in each of the samples used in this study.
*Pfkelch13* PCR sequencing strategy {#Sec7}
-----------------------------------
The propeller domain of *pfkelch13* was PCR amplified using Phusion HF DNA Polymerase kit and primers 3F and 1R′ in all 68 culture adapted parasites. DNA from KH001_024 was also amplified with primers 4F and 3R′. An aliquot of PCR product was resolved by gel electrophoresis to check for specificity and yield and the remaining product was purified using DNA Clean & Concentrator, ZymoResearch and sequenced using the same primers used to amplify the product by Genewiz. The full ORF of *pfkelch13* was PCR amplified using primers 1F and 1R from 3D7 and individual culture-adapted parasites. The resulting PCR product of \~2.2 kb was purified by gel extraction (QIAquick Gel Extraction Kit, Qiagen) and sequenced at Genewiz using primers 1F, 2F, 3F, 1R, 2R and 3R. Primer sequences are as follows: 1F: 5′-ATGGAAGGAGAAAAAGTAAAAACAAAAGCAAATAG-3′; 2F: 5′-GGTAGGTGATTTAAGAATTACATTTATTAATTGGT-3′; 3F: 5′-CATTCCCATTAGTATTTTGTATAGGTG-3′; 4F: 5′-GTAGAGGTGGCACCTTTGAATACCCCTAGATCATC-3′ 1R: 5′-TTATATATTTGCTATTAAAACGGAGTGACCAAATCTG-3′; 1R′: 5′-TTA TAT ATT TGC TAT TAA AAC GGA GTG-3′; 2R: 5′-AGCCTTATAATCATAGTTATTACCACCAAAAACG-3′; 3R: 5′-TGTTGGTATTCATAATTGATGGAGAATTC-3′; 3R′: 5′-ATAAAATGTGCATGAAAATAAATATTAAAG-3′.
In vitro 72-h drug susceptibility by SYBR green staining {#Sec8}
--------------------------------------------------------
Drug susceptibility was measured using the SYBR Green I method as previously described \[[@CR27]\]. Briefly, tightly synchronized 0--6 h rings were grown for 72 h in the presence of different concentrations of drugs in 384-well plates at 1% haematocrit and 1% starting parasitaemia; and, growth at 72 h was measured by SYBR Green staining of parasite DNA. Except for PPQ and KH001_053 where a 24-point dilution series was used, a 12-point dilution series of each drug was carried out in triplicate and repeated with at least three biological replicates. DMSO stocks of drugs were dispensed by a HP D300 Digital Dispenser (Hewlett Packard Palo Alto, CA, USA) except for the CQ and PPQ stocks that were prepared in water and dispensed with a Velocity 11 Robot (Bravo). Relative fluorescence units (RFU) was measured at an excitation of 494 nm and emission of 530 nm on a SpectraMax M5 (Molecular Devices Sunnyvale, CA, USA) and analysed using GraphPad Prism version 5 (GraphPad Software La Jolla, CA. USA). EC~50~ values were determined with the curve-fitting algorithm log(inhibitor) vs response---variable slope, except for PPQ and KH001_053. Due to the bimodal dose response of KH001_053 to PPQ, curve fitting didn't give an accurate EC~50~ value. The reported PPQ EC~50~s for KH001_053 are estimates using biphasic curve fitting. Spearman correlation analysis was performed to assess the relationship between the anti-malarial EC~50~ values and in vivo clearance half-life, ring survival assay value or *pfmdr1* copy number. p values \<0.05 were considered significant.
Copy number variation assays {#Sec9}
----------------------------
To determine copy numbers for *pfmdr1, plasmepsin II* and the 63 kb amplicon genes (PF3D7_0520100, PF3D7_0520500, PF3D7_0520600, PF3D7_0520900 and PF3D7_0521000), real time quantitative PCR was performed on genomic DNA (extracted with QIAmp Blood Mini Kit, Qiagen) as previously described \[[@CR28]\] with the following modifications: Amplification reactions were done in MicroAmp 384-well plates in 10 μL SYBR Green master mix (Applied Biosystems), 150 nM of each forward and reverse primer and 0.4 ng template. Forty cycles were performed in the Applied Biosystems ViiA™ 7 Real-time PCR system (Life Technologies). Forward and reverse primers used were as previously described to amplify the following loci: *pfmdr1* (PF3D7_0523000) \[[@CR29]\], the 63 kb region on chromosome 5 (PF3D7_0520100, PF3D7_0520500, PF3D7_0520600, PF3D7_0520900 and PF3D7_0521000 \[[@CR20]\]) and *plasmepsin II* (PF3D7_140800) \[[@CR22]\]. For the endogenous controls, *β*-*tubulin* forward and reverse primers \[[@CR28]\] were used for *pfmdr1,* PF3D7_0520100 and PF3D7_0520900 while *pfldh* forward and reverse primers \[[@CR30]\] were used for PF3D7_0520500, PF3D7_0520600 and PF3D7_0521000. Target primers used were validated to have the same PCR efficiencies as their endogenous control primers; and, average copy number values were calculated for each gene using data from three independent experiments.
Sequencing the *pfcrt* locus {#Sec10}
----------------------------
The entire *pfcrt* locus was sequenced as previously described \[[@CR20]\] with some modifications. Briefly, total RNA was extracted using RNeasy kit (Qiagen) and used to generate cDNA using Superscript III (Invitrogen). The resulting cDNA was then used as template for PCR amplification of *pfcrt* \[[@CR20]\], followed by Sanger sequencing (GENEWIZ) \[[@CR20]\]. Sequence data analysis was performed using MacVector.
Ring survival assay (RSA~0--3h~) {#Sec11}
--------------------------------
The RSA~0--3h~ was performed as described previously \[[@CR6]\]. Essentially, parasites were sorbitol synchronized twice at 40-h intervals, synchronous 40--44 h segmented schizonts were incubated for 15 min at 37 °C in serum-free media supplemented with heparin to disrupt agglutinated erythrocytes and late stages were purified with 35/65% discontinuous Percoll gradient. The segmented schizonts were washed and cultured with fresh RBCs for 3 h, after which late stages were removed by sorbitol treatment. Cultures with 0--3 h rings were adjusted to 2% haematocrit and 1% parasitaemia and seeded into a 24-well plate with 1 ml complete media per well. To these wells, either DHA at 700 nM or 0.1% DMSO were added immediately and incubated for 6 h at 37 °C, washed and incubated in drug free media. At 72 h from seeding, thin blood smears were made from control and treated wells and survival rates were measured microscopically by counting the proportion of next generation viable rings with normal morphology. Survival rates were expressed as ratios of viable parasitaemias in DHA-exposed and DMSO-treated controls. Parasites were counted from 10,000 RBC, and two separate individuals served as independent slide readers.
Results {#Sec12}
=======
Characteristics of original and culture-adapted parasite isolates {#Sec13}
-----------------------------------------------------------------
A total of 157 cryopreserved TRAC study samples collected in 2011 from Pursat or Pailin, Cambodia where ART resistance is observed were obtained. These isolates were selected before the discovery of the *pfkelch13* marker and were designed to comprise western Cambodian samples, with two-thirds of the samples coming from the upper segment of the clearance half-life distribution and one-third being isolates with the shortest clearance times. 68 parasites were culture-adapted for further investigation, without knowledge of their in vivo clearance data. A summary of the overall population and the sub-set selected for culture-adaptation and further characterization is provided in Additional files [2](#MOESM2){ref-type="media"} and [3](#MOESM3){ref-type="media"}. Collectively, 63% of both the original and the adapted parasites were from Pursat and the remaining 37% were from Pailin (Additional file [3](#MOESM3){ref-type="media"}). Similarly, 67% of both parasite sets (original and adapted) had a delayed clearance of ≥5 h and 33% had a clearance half-life of \<5 h; for one isolate there were insufficient data to determine an in vivo half-life. All culture-adapted samples were confirmed to harbour monogenomic infections by molecular barcode analysis \[[@CR25]\], only those parasites with unique barcode signatures \[[@CR25]\] were initially pursued to maximize genetic diversity and exclude highly similar parasites among the analysed isolates.
*Pfkelch13* propeller mutations are associated with parasite clearance half-life ≥5 h {#Sec14}
-------------------------------------------------------------------------------------
Next, parasites were genotyped for the *pfkelch13* locus previously associated with delayed parasite clearance (defined as \>5 h) and increased ring-stage survival (defined as RSA~0--3h~ ≥1%) \[[@CR7]\]. An HRM genotyping assay \[[@CR24], [@CR25]\] for the common non-synonymous mutation conferring the C580Y change in *Pf*Kelch13 was developed and validated, among the 157 original samples with an unambiguous genotyping call, 71% (108/152) had the C580Y mutation (Additional file [2](#MOESM2){ref-type="media"}). These data, confirmed a significant positive association between the C580Y non-synonymous mutation and in vivo parasite clearance half-life values for these parasites \[[@CR26]\]. This association was more pronounced among Pursat parasites (unpaired Student's t test with Welch's correction, p \< 0.0001) as compared to parasites from Pailin (p = 0.0322, Additional file [4](#MOESM4){ref-type="media"}) for this sample set.
A combination of available whole genome sequence (WGS) data from the original samples \[[@CR26], [@CR31], [@CR32]\], and *pfkelch13*-specific PCR-based sequencing identified both *pfkelch13* mutations other than C580Y in this population and confirmed persistence of all *pfkelch13* mutations in the culture adapted parasites further analysed. Overall there was good concordance between the molecular barcode derived from filter paper and cultured parasite samples, and between the *pfkelch13* mutations identified in WGS or PCR re-sequencing data. Six different mutations were present in the *pfkelch13* propeller domain: Y493H, R539T, I543T, C580Y, D584V, and H719N, and two *pfkelch13* mutations outside the propeller region: H136N, E270K. The two parasites with the H136N change also contained C580Y, but E270K was seen by itself, in a single isolate (Additional file [2](#MOESM2){ref-type="media"}). All propeller domain mutations were positively associated with in vivo clearance values (Fig. [1](#Fig1){ref-type="fig"}a), including parasites with the double mutation H136N/C580Y genotype (Additional file [2](#MOESM2){ref-type="media"}). Additionally, two asparagine (Asn) insertions at codon 142 were identified in all isolates tested except those with the Y493H change and one parasite with the D584V genotype (Additional file [2](#MOESM2){ref-type="media"}). The location of this Asn insertion was distinct from that previously noted among Asian parasites \[[@CR33]\]. No association between this Asn insertion and either parasite clearance half-life or RSA~0--3h~ phenotype was detected.Fig. 1Positive association between *pfkelch13* propeller mutations and in vivo clearance half-life as well as in vitro RSA~0--3h.~ Parasites from Pursat (*round symbols*) or Pailin (*square symbols*) were classified according to their *pfkelch13* alleles. **a** Comparison of clearance half-live values (hours, h) from culture-adapted parasites harbouring different *pfkelch13* propeller alleles, and combining alleles represented by less than five samples. Parasites outlined in bold were further tested for RSA~0--3h~ phenotypes in **b**. A one-way ANOVA test, with a Tukey's post-test, was performed between wild-type (WT) and each allele category. Significance values are indicated by *asterisks*: \*(p \< 0.05); \*\*(p \< 0.01); \*\*\*(p \< 0.001); and \*\*\*\*(p \< 0.0001). **b** In vitro RSA~0--3h~ testing. Percent survival is displayed on the y-axis and *pfkelch13* allele represented on the x-axis. All parasites that harbour a *pfkelch13* propeller mutation exhibit an RSA~0--3h~ value of \>1% (*dotted line*). Data were analysed by Kruskal--Wallis test followed by Dunn's multi-comparison test. p values: \*\*\*\<0.001. An additional subset represented by *coloured symbols*, also underwent conventional in vitro drug testing and assessment of *pfmdr1* copy number variation. These were chosen as representatives of parasites that were ART sensitive (*green*) or, ART resistant (*red*) as assessed by both RSA~0--3h~ phenotype and *pfkelch13*. Two 'discordant' samples (resistant according to RSA but wild-type at *pfkelch13*) are indicated in *blue*
The sub-set of 68 culture-adapted parasites harbored all mutation types observed in the 157 initial isolates except H136N, and each *pfkelch13* mutation detected by WGS was confirmed by PCR re-sequencing (Additional files [2](#MOESM2){ref-type="media"}, [3](#MOESM3){ref-type="media"}). Overall, the sub-set of culture-adapted parasites showed similar distribution of *pfkelch13* genotypes detected among the original 157 isolates chosen for this study. The majority of parasites in this sub-set (79% or 54/68) carried non-synonymous mutations in *pfkelch13,* with the most common change resulting in C580Y (53% or 36/68, Table [1](#Tab1){ref-type="table"}). Among these culture-adapted parasites from Pailin there was a higher frequency of *pfkelch13* mutations than parasites from Pursat (88 and 74%, respectively), with the C580Y mutation being the most prevalent (77 and 60%, respectively). Both geographic locations had their own private mutations not found in the other location (Additional file [2](#MOESM2){ref-type="media"}). All *pfkelch13* propeller mutations displayed positive associations with in vivo clearance phenotypes (Fig. [1](#Fig1){ref-type="fig"}a).Table 1Summary of *Pf*Kelch13 mutationsCulture adapted parasitesTotalWTY493HR539TI543TC580YD584VH719NH136N/C580YE270KAsn InsertPailin2530221700016Pursat431156019110019Total681458236110125
*Pfkelch13* propeller mutations are positively associated with increased ring-stage survival by in vitro RSA~0--3h~ {#Sec15}
-------------------------------------------------------------------------------------------------------------------
There was a positive association between increased RSA~0--3h~ survival phenotypes and *pfkelch13* mutations among 36 culture adapted parasites (Fig. [1](#Fig1){ref-type="fig"}b) and parasites with the D584V change previously associated with increased in vivo parasite clearance time \[[@CR5]\] demonstrated an increased ring-stage survival by RSA~0--3h~.
In vitro RSA~0--3h~ phenotypes were evaluated as a measure of ART resistance for 36 adapted parasites exhibiting a range of in vivo clearance phenotypes and harbouring a variety of *pfkelch13* mutations. Consistent with previous reports \[[@CR6]\], the distribution of RSA~0--3h~ phenotypes spanned two log~10~ ranges for all parasites analysed. Overall, there was a significant difference between wild-type (RSA~0--3h~ = 0.1--6.0%, n = 7) parasites and those with *pfkelch13* mutations (RSA~0--3h~ = 2.1--38.6%, n = 29) in terms of RSA~0--3h~ parasite survival rate (Student's t test p = 0.0018). RSA~0--3h~ survival rates also differed between parasites with individual *pfkelch13* mutations; whereby parasites with either a C580Y or R539T mutation exhibited the highest survival rates (Fig. [1](#Fig1){ref-type="fig"}b). One isolate with a *Pf*Kelch13 E270K mutation outside the propeller domain exhibited a clearance of \>4 h, yet had an RSA~0--3h~ value of 0.4%. Among the parasites with RSA~0--3h~ values \>1%, 19% (7/36) had in vivo clearance values between 4 and 5 h, which suggests that for this specific population a clearance value of 4 h or greater is more consistent with ART resistance based upon an RSA~0--3h~ phenotype (Additional file [5](#MOESM5){ref-type="media"}).
All *pfkelch13* propeller polymorphisms found among these parasites were previously identified \[[@CR7], [@CR33]\]. However, a novel positive association was demonstrated between the D584V mutation and in vitro ring survival values (8.0 h clearance and 8.5% RSA survival for D584V), not previously tested for in vitro ART resistance (Fig. [1](#Fig1){ref-type="fig"}b). These data confirm that all *pfkelch13* propeller mutations found in these TRAC isolates are associated with an RSA~0--3h~ phenotype ≥1%, thus considered in vitro ART resistant.
Identification of discordant parasites that exhibit increased ring-stage survival but lack *Pfkelch13* mutations {#Sec16}
----------------------------------------------------------------------------------------------------------------
ART resistance in vitro has been suggested to correlate with an RSA~0--3h~ survival value of ≥1% \[[@CR6]\], and all isolates in this study with a mutation in *pfkelch13* had an RSA~0--3h~ ≥1%. All wild-type *pfkelch13* parasites had in vivo clearance half-lives below 5 h (Additional file [2](#MOESM2){ref-type="media"}); however, there was a range of in vitro RSA~0--3h~ survival phenotypes from 0.1 to 6%, including four isolates with RSA~0--3h~ of approximately ≥1% (0.8, 1.1, 3, and 6%) (Fig. [1](#Fig1){ref-type="fig"}b). Parasites with an RSA~0--3h~ of ≥0.8% that lack *pfkelch13* mutations are considered discordant. These data suggest changes outside *pfkelch13* could also confer an increased ring-stage survival phenotype indicative of ART resistance in this parasite population.
Using WGS data previously generated \[[@CR26]\] for 33 of the RSA~0--3h~ phenotyped parasites, the question was explored whether mutations outside *pfkelch13* were evident in the four parasites with RSA~0--3h~ ≥0.8% values that lack *pfkelch13* ORF mutations. This small data set was underpowered to identify novel mutations using a genome-wide strategy, but a candidate gene approach was undertaken to ask whether any known drug resistant loci or other previously identified secondary mutations could potentially explain the high RSA~0--3h~ survival values (Additional file [6](#MOESM6){ref-type="media"}). Variant positions were identified across 23 genes previously associated with ART resistance among the set of 33 RSA~0--3h~ phenotyped parasites \[[@CR7], [@CR15], [@CR26]\], which included both sensitive and resistant parasites. Examination of both sensitive and resistant parasites would allow detection of mutations that may account for the differences in RSA~0--3h~ survival values and identification of mutations that confer high RSA~0--3h~ survival values independent of *pfkelch13*.
This analysis identified 37 SNPs (30 non-synonymous, four synonymous, two intronic) spread across 19 unique genes (Additional file [7](#MOESM7){ref-type="media"}). Eleven of these positions were invariant among these four isolates that lack *pfkelch13* ORF mutations but exhibited an RSA~0--3h~ survival value ≥0.8%. These 37 SNP positions are found in genes known to be associated with drug resistance including dihydrofolate reductase (*dhfr*, PF3D7_0417200) \[[@CR34]\] and dihydropteroate synthase (*dhps*, PF3D7_0810800) \[[@CR35]\]. Another locus, phosphatidylinositol-4-phosphate 5-kinase (PIP5K, PF3D7_0110600) \[[@CR36]\], is involved in cellular signaling pathways, including synthesis of PIP2, a substrate for PI3K to produce PIP3, which activates the AKT family of serine/threonine kinases, a pathway implicated in an ART resistance mechanism \[[@CR15]\]. The putative NLI Interacting factor-like phosphatase (NIF4, PF3D7_1012700) \[[@CR36]--[@CR39]\] is a protein phosphatase 2C protein. Several members of the *Plasmodium* phosphatome contain *kelch* domains and have been shown to be key regulators of parasite development and differentiation \[[@CR37], [@CR40]\], but NIF4 lacks an intact DxDx(T/V) motif, and thus may be catalytically inactive \[[@CR39]\]. *Pfmdr2* (PF3D7_1447900) \[[@CR41], [@CR42]\] is implicated as a secondary locus associated with delayed parasite clearance \[[@CR26]\]. No other changes associated with ART resistance reported in the literature \[[@CR43], [@CR44]\] were detected among these parasites (Additional file [6](#MOESM6){ref-type="media"}).
Detection of PPQ resistance phenotype {#Sec17}
-------------------------------------
To further evaluate the relationship between RSA~0--3h~ phenotype and *pfkelch13* mutations, a sub-set of eight isolates with a range of ring-stage survival phenotypes was selected. This sub-set included the two most sensitive (in green) and the two most resistant (discordant, in blue) parasites without *pfkelch13* mutations, as well as four resistant parasites that contain *pfkelch13* mutations (in red) (Fig. [1](#Fig1){ref-type="fig"}b). First standard in vitro drug tests \[[@CR27]\] were used to evaluate anti-malarial drug responses among these parasites (Fig. [2](#Fig2){ref-type="fig"}). EC~50~ values for ART and its derivatives did not consistently correlate with either the in vivo clearance or in vitro RSA~0--3h~ phenotypes, as previously reported \[[@CR6], [@CR7]\] (Additional file [8](#MOESM8){ref-type="media"}; Fig. [2](#Fig2){ref-type="fig"}). The parasite clearance half-life and RSA~0--3h~ survival phenotypes did not consistently associate with responses to other anti-malarial compounds tested (chloroquine (CQ), MFQ, lumefantrine (LUM), atovaquone (ATV), quinine (QN) and PPQ, Additional file [8](#MOESM8){ref-type="media"}). A significant positive correlation was observed between EC~50~ values for ART (r = 0.881, p = 0.0072), artesunate (AS, r = 0.719, p = 0.0368), MFQ (r = 0.7619, p = 0.0368) and LUM (r = 0.857, p = 0.0107) and increased *pfmdr1* copy number (Spearman correlation, Additional file [8](#MOESM8){ref-type="media"} and Fig. [2](#Fig2){ref-type="fig"}), consistent with previous reports \[[@CR30], [@CR45], [@CR46]\].Fig. 2*Pfmdr1* copy numbers are positively associated with MEF, LUM, ART and ARS. Drug sensitivity assays were set up with 0--6 h old ring stage parasites and % growth was measured 72 h later by SYBR *Green*. Shown are the average EC~50~ values from at least three biological replicates for two ART sensitive (*green*), two discordant (*blue*) and four resistant (*red*) isolates. Their respective clearance time, RSA~0--3h~ percent survival and their *pfmdr1* copy number are indicated *at the bottom*. The Spearman correlation coefficient for *pfmdr1* copy number and each drug is shown in the graph. A one-way ANOVA test, with a Tukey's post-test was performed between all isolates for each drug. Significance values are indicated by *asterisks*, as follows: \*(p \< 0.05); \*\*(p \< 0.01)
Interestingly, one isolate, KH001_053, with a slow clearance time (8.7 h) and high ring survival (30%) exhibited a bimodal response to PPQ that indicated parasite survival at high PPQ concentrations (Fig. [3](#Fig3){ref-type="fig"}a). Compared to other isolates, KH001_053 also demonstrated a high survival phenotype (50--70%) when grown at high concentrations of PPQ (125 and 250 nM) for 72 h \[[@CR47]\] (Fig. [3](#Fig3){ref-type="fig"}b). This PPQ-resistant parasite also demonstrated hyper-susceptibility to MFQ and LUM (Fig. [2](#Fig2){ref-type="fig"}), explicable at least in part by a single copy of the *pfmdr1* gene (Additional file [8](#MOESM8){ref-type="media"}).Fig. 3KH001_053 exhibits resistance to piperaquine (PPQ). **a** EC~50~ growth curves for eight parasite isolates in different PPQ concentrations. Percent growth is displayed on the y-axis and the PPQ concentration (nM) is represented in a log scale on the x-axis, with KH001_053 indicated by *red triangles*. **b** Relative growth is shown (percent growth) for TRAC isolates exposed to 125 nM PPQ for 72 h starting with 0--6 h old rings, with individual parasite lines represented on the x-axis. *Colours* correspond to parasites with RSA~0--3h~ ≤0.2% that lack *pfkelch13* mutations (sensitive, *green*); RSA~0--3h~ \>3% that lack *pfkelch13* mutations (discordant, *blue*); and, RSA~0--3h~ \>1% that harbour with *pfkelch13* mutations (resistant, *red*). Data were analysed by Kruskal--Wallis test followed by Dunn's multi-comparison test. p values: \*\<0.1, \*\*\<0.01, \*\*\*\*\<0.0001
KH001_053 was then tested for evidence of known molecular markers associated with PPQ resistance \[[@CR20]\], including C101F mutation in *pfcrt*, an amplification of 63 kb region on chromosome 5, and amplification of *plasmepsin II* \[[@CR21], [@CR22]\]. All eight isolates were evaluated for SNPs in the *pfcrt* locus by PCR amplification and sequencing of cDNA across *pfcrt*. Eight non-synonymous mutations (M74I, N75E, K76T, A220S, Q271E, N326S, I356T, R371I) were found that were fixed in these eight isolates. No evidence of the previously reported C101F haplotype in the KH001_053 parasite was detected \[[@CR20]\]. Since PPQ resistance has been associated with an increased copy number of a 63 kb amplicon on chromosome 5 (including PF3D7_0520100, PF3D7_0520500, PF3D7_0520600, PF3D7_0520900 and PF3D7_0521000, 825600--888300 bp) and a decreased number of *pfmdr1* in vitro \[[@CR20], [@CR30]\] as well as an increased copy number of *plasmepsin II* (PF3D7_140800) locus in vivo \[[@CR21], [@CR22]\], copy number assays were performed on those loci. In concordance with to a previous study \[[@CR30]\] no increased copy number on the 63 kb amplicon region on chromosome 5 was detected. However, KH001_053 did harbour two copies of *plasmepsin II* consistent with previous reports where the presence of an increased *plasmepsin II* and *III* copy number was predictive of recrudescence in DHA/PPQ-treated patients \[[@CR21], [@CR22]\]. These findings strengthen the hypothesis that increased *plasmepsin II* copy number may play a role in PPQ resistance.
Discussion {#Sec18}
==========
Within a set of 68 culture-adapted Cambodian parasites, some parasites exhibited an increased RSA~0--3h~ phenotype that lack *pfkelch13* mutations, a parasite harbouring the *Pf*Kelch13 D584V change exhibited increased RSA~0--3h~ survival, and some parasites showed evidence for PPQ resistance. To evaluate a set of 157 Cambodian *P. falciparum* isolates from the TRAC study, with known clearance half-life data, a genotyping assay for the common C580Y change associated with ART resistance was developed and confirmed a clear association between delayed parasite clearance half-life and the presence of the C580Y allele. Sequence or genotype data from 146 of these parasites demonstrated a positive association between the clearance phenotype and *pfkelch13* propeller mutations. 68 of these parasites were culture-adapted and 36 were tested for their in vitro RSA~0--3h~ phenotype. Increased parasite clearance half-life was associated with RSA~0--3h~ ≥1% values, with a 4-h or greater clearance phenotype being more consistent with ART resistance based upon RSA~0--3h~ for this population. A novel positive association was demonstrated between D584V and in vitro ring-stage parasite survival under DHA among these parasites.
Despite this general correspondence between in vivo clearance and in vitro RSA~0--3h~ phenotypes and *Pfkelch13* propeller mutations, there were discordant parasites that exhibited a resistant RSA~0--3h~ phenotype (RSA~0--3h~ ≥0.8%) yet harboured no mutations in the entire *pfkelch13* ORF, consistent with other studies that found similarly discordant parasites by RSA~0--3h~ \[[@CR48]\], or by clearance half-life \[[@CR26]\]. A parasite (see Additional file [5](#MOESM5){ref-type="media"}) with a *Pf*Kelch13 E270K outside the propeller domain exhibited a clearance value of \>4 h, yet had an RSA~0--3h~ value of 0.4 %. These differences may be due to differential host factors including variances in drug pharmacokinetics, pharmacodynamics or host immune status. More isolates with an E270K mutation need to be studied to confirm the role of a *pfkelch13* mutation outside the propeller domain in ART resistance and reduce the probability of an artifact related to microscopy. Taken together, these results suggest that ART resistance in natural parasite populations could be mediated by changes outside of the *pkelch13* propeller domain. Parasites with these phenotypes around the defined cut off values may be important for identification and study of loci other than *pfkelch13* that contribute to ART resistance. Inspection of candidate loci, including traditional drug resistance mutations, recently identified secondary mutations, or loci related to pathways implicated in ART resistance, identified a few specific changes that may account for these discordant parasites, but lacked the statistical power to determine whether any of these contributed to ART resistance. However, identification of potential mutations in the *PIP5K* and *pfmdr2* loci is consistent with other reports. Nevertheless, the identification of parasites that lack *pfkelch13* mutations yet harbour an increased RSA~0--3h~ survival phenotype, suggest loci other than *pfkelch13* may modulate ART resistance in these parasites. Alternative strategies, such as the use of independent chemogenomic strategies or genetic crosses in vitro \[[@CR49]\], might be useful to identify loci involved in conferring the observed increased ring-stage survival phenotype in the absence of *pfkelch13* mutations.
No relationship was found between EC~50~ values for ART or its derivatives with either clearance or ring-stage survival phenotypes, but a previously noted correspondence was confirmed between *pfmdr1* copy number and EC~50~ values of ART and AS, as well as with MFQ and LUM responses, among eight parasites evaluated for drug responses (Fig. [2](#Fig2){ref-type="fig"}). Drug testing identified a PPQ-resistant isolate among this population, consistent with reports in the literature of increased PPQ resistance in this region \[[@CR18], [@CR19]\]. The PPQ-resistant parasite had a single copy of the *pfmdr1* gene \[[@CR20], [@CR30]\] and amplification of the *plasmepsin II* locus previously noted \[[@CR21], [@CR22]\]. The exact nature of this resistance, the role of *plasmepsin II* and the explanation for the apparent bimodal response to PPQ remains unknown and under investigation, but identification of culture-adapted parasites affords additional testing of both the phenotype and genotype (Additional file [1](#MOESM1){ref-type="media"}).
It has been suggested \[[@CR18], [@CR19]\] that PPQ resistance arose in the context of or as a consequence of ART resistance. Understanding the nature of this resistance and the underlying mechanism will be critical for reducing or restricting emergence of PPQ resistance in Southeast Asia where reports indicate rapid emergence over the past several years \[[@CR18]\]. Furthermore, since DHA--PPQ is being utilized in mass drug administration campaigns designed to facilitate elimination of malaria in specific settings, use of molecular markers of resistance and understanding the relationship between ART and PPQ resistance will be important. Thus, identification of PPQ resistance with MFQ and LUM hyper-susceptibility in an ART resistant parasite could be an important clue to understanding mechanisms of drug response, and testing the implications of triple therapy, such as ART--PPQ--MFQ, being considered for use in this region \[[@CR18]\].
Conclusions {#Sec19}
===========
In a large set of *P. falciparum* isolates from the TRAC study, the associations between C580Y and several other *pfkelch13* propeller mutations and parasite clearance half-life was investigated. In a subset of 68 culture-adapted parasites, RSA~0--3h~ survival and conventional responses to multiple antimalarial drugs were measured. Several *pfkelch13* mutations (including D584V) were associated with increased RSA~0--3h~ survival, and discordant parasites with RSA~0--3h~ survival 1% but without *pfkelch13* ORF mutations were identified. These data suggest that mutations outside of *Pfkelch13* may confer in vitro ART resistance in *P. falciparum*. It will therefore be important to continue phenotypic assessment of ART resistance, in addition to surveying for *pfkelch13* propeller mutations. Detection of a PPQ-resistant parasite will enable further studies to investigate underlying mechanisms of PPQ resistance. This panel of culture-adapted parasites with known parasite clearance half-life, RSA~0--3h~ survival, and *pfkelch13* genotype will facilitate further investigation of ART resistance mechanisms, providing tools to identify potential *Pf*Kelch13-binding partners and other interacting molecules.
Additional files
================
{#Sec20}
**Additional file 1.** Culture-adaptation of TRAC parasites. **Additional file 2.** Summary of all genotypes available. **Additional file 3.** Schematic shows distribution of parasites obtained from Pursat (blue star on map, KH1 parasites indicated in blue) and Pailin (red star, KH4 parasites indicated in red) from Cambodia (map modified from <https://commons.wikimedia.org/wiki/File:Cambodia_provinces_en.svg>). **Additional file 4.** Distribution of C580Y mutations in Cambodian isolates. Parasites from Pursat (round symbols) or Pailin (square symbols) were classified according to their *Pfkelch13* alleles. Comparison of clearance half-life values (hours, h) from parasites harboring the wild-type (C) or mutant (Y) allele at amino acid position 580 in the Pfkelch13 locus, with culture-adapted parasites represented by black outlined symbols. C580 parasites harboring a mutation in the Pfkelch13 propeller domain other than C580Y are indicated by grey filled symbols. An unpaired t test was performed between the C580 and C580Y carrying parasites from Pursat (p \< 0.0001) or Pailin (p \< 0.05); with significant differences (p \< 0.0001 overall) when both Pursat and Pailin were combined. **Additional file 5.** Association between percent survival RSA~0--3h~ and clearance time. Correlation between clearance time on the x-axis and percent survival RSA~0--3h~ on the y-axis. Each dot represents an isolate. Samples represented by color were used for in vitro drug testing and for *pfmdr1* copy number variation determination---green for ART sensitive; blue for discordant; and red for ART resistant parasites. The grey dot marks the isolate with an E270K mutation. The dotted lines represent the cut off used to discriminate between resistant and sensitive parasites (1% for RSA~0--3h~, 4 or 5% for clearance). **Additional file 6.** All genes analysed for candidate gene approach. **Additional file 7.** All positions that are variant in RSA~0--3h~ phenotyped isolates. **Additional file 8.** Spearman correlation coefficients between antimalarial EC50 values, in vivo clearance half-life, ring survival assay~0--3h~ or pfmdr1 copy number.
ACT
: artemisinin combination therapy
ART
: artemisinin
AS
: artesunate
Asn
: asparagine
ATV
: atovaquone
CQ
: chloroquine
DHA
: dihydroartemisinin
HRM
: high resolution melt
iRBC
: infected red blood cell
LUM
: lumefantrine
MFQ
: mefloquine
NIF4
: NLI interacting factor-like phosphatase
ORF
: open reading frame
PCR
: polymerase chain reaction
PPQ
: piperaquine
QN
: quinine
RBC
: red blood cells
SNP
: single nucleotide polymorphism
TRAC
: Tracking Resistance to Artemisinin Collaboration
WGS
: whole genome sequencing
RSA~0--3h~
: ring-stage survival assay
PIP5K
: phosphatidylinositol-4-phosphate 5-kinase
CNV
: copy number variation
**Electronic supplementary material**
The online version of this article (doi:10.1186/s12936-017-1845-5) contains supplementary material, which is available to authorized users.
Angana Mukherjee and Selina Bopp contributed equally to this work
AM, SB, PM carried out RSA~0--3h~, drug testing, PCR-based genotyping and copy number variation analysis and helped write the manuscript; WW carried out the genomic analysis with data provided by OM and helped write the manuscript; RD developed the C580Y HRM genotyping assay and carried out molecular barcoding of the samples; AD created primer sets and helped with RSA~0--3h~ analysis of the samples; SS helped with genomic analysis and provided critical edits to the manuscript; CA, PL, MD, CW, EAA, AMD, NJW, and RF carried out field collection and provided in vivo sample information and samples. DFW provided critical review of the data and experimental guidance. SKV culture adapted the parasites, provided critical review of the data and experimental guidance, and wrote the manuscript. All authors provided critical review of the data and manuscript before publication. All authors read and approved the final manuscript.
Acknowledgements {#FPar1}
================
We would like to thank the patients who participated to the TRAC study and Dr. Kesinee Chotivanich and her team for initial cryopreservation of the isolates. We would like to thank Courtney Edison for help with tissue culture and RSA assays, and Elizabeth Hamilton for work developing the C580Y genotyping assay and for genotyping filter paper derived samples. We would also like to thank Katelyn Durfee for molecular barcode analysis of culture-adapted lines. We thank Paul Hinkson for his help with EC~50~ assays for anti-malarial drug testing of the culture-adapted parasites.
Competing interests {#FPar2}
===================
The authors declare that they have no competing interests.
Availability of data and materials {#FPar3}
==================================
Datasets generated and/or analysed during the current study are available through the MalariaGEN Pf3K Project \[[@CR26], [@CR32]\].
Ethics approval and consent to participate {#FPar4}
==========================================
The scientific merit and use of human subjects for this study was approved by the institutional review board (IRB) of the Harvard T. H. Chan School of Public Health (22689-101). Written informed consent was obtained from all adult subjects and from the parent or legal guardians of minors.
Funding {#FPar5}
=======
We thank the generous funding that supported this work. The TRAC study was funded by the UK Department for International Development; support for culture-adaptation and analysis at the Harvard T. H. Chan School of Public Health was provided by the University of Oxford/Bill and Melinda Gates Foundation for "Identifying a Laboratory Marker of Artemisinin Resistance" (OPP104063, PI: Nicholas White); and from the Bill and Melinda Gates Foundation "Genomics-Based Diagnostics for Elimination and Eradication of *Plasmodium*" (OPP1053604, PI: Dyann Wirth).
Publisher's Note {#FPar6}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-cancers-12-00185}
===============
Diffuse large B-cell lymphoma (DLBCL) comprises about 40% of all lymphomas, constituting the most prevalent type of non-Hodgkin lymphoma. It can arise de novo or result from the malignant transformation of a more indolent lymphoma. According to *The 2016 revision of the World Health Organization (WHO) classification of lymphoid neoplasms* there are 13 subtypes of lymphoma defined as specific entities, designating the rest as DLBCL not otherwise specified (NOS), which account for the vast majority of DLBCLs \[[@B1-cancers-12-00185]\]. The standard treatment approach consists of immunochemotherapy (rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone---R-CHOP), which guarantees an overall survival (OS) of more than 60% for DLBCL-NOS cases. In particular, a subgroup of young patients with favourable-prognosis disease can even achieve the same clinical benefit with fewer cycles of R-CHOP \[[@B2-cancers-12-00185]\]. However, up to 40% of patients suffer relapse or refractory (R/R) disease \[[@B3-cancers-12-00185]\] and for them the standard salvage approach consists of autologous stem cell transplantation, even if long-term disease control is achieved in fewer than 50% of cases \[[@B4-cancers-12-00185]\]. Survival is particularly poor for patients relapsing within one year after R-CHOP with fewer than 15% of patients achieving a durable remission \[[@B5-cancers-12-00185],[@B6-cancers-12-00185],[@B7-cancers-12-00185]\]. Recently chimeric antigen receptor (CAR) T-cell therapies have been approved as alternative curative options for patients with relapsing or refractory disease. CAR T cells represent a new class of cellular immunotherapy involving ex vivo genetic modification of patients' T cells, triggering T-cell activation and cytotoxicity \[[@B8-cancers-12-00185]\], that demonstrated good efficacy in B-cell malignancies treatment, including DLBCL \[[@B9-cancers-12-00185],[@B10-cancers-12-00185]\]. In this context, a prevision of poor OS is attributed to relapsing cases and to patients with refractory disease \[[@B6-cancers-12-00185]\] for which even CAR T-cell therapy fails \[[@B11-cancers-12-00185]\].
Therefore, it is essential to search for clinical parameters and biomarkers that could help to better DLBCL patients' characterization and stratification. Nowadays, thanks to the availability of comprehensive genomic and transcriptomic analyses a wealth of information is generated, rendering the concept of personalized therapy more realistic. In the attempt to put some order in the most recent discoveries on DLBCL research, we reviewed the latest experimental studies in this field, focusing on the most important findings helping in the management of lymphoma patients from the perspective of personalized medicine.
1.1. Standard Prognosticators for DLBCL {#sec1dot1-cancers-12-00185}
---------------------------------------
One of the most commonly used prognostic tools is the *International Prognostic Index* (IPI) \[[@B12-cancers-12-00185]\], whose validity and reliability has been enhanced by several upgrades \[[@B13-cancers-12-00185]\]. However, it evaluates only five clinical parameters (age, lactate dehydrogenase, performance status, number of extranodal sites, and Ann Arbor stage), without considering the biologic characteristics of the tumour.
The first and nowadays most commonly used biologic prognosticator of DLBCL tumours is the cell-of-origin (COO) determination based on gene expression profiling (GEP), which subdivides most DLBCL-NOS patients into two main categories, namely germinal center B-cell-like (GCB), if presenting with expression features similar to germinal center cells, and activated B-cell-like (ABC) DLBCL \[[@B14-cancers-12-00185]\], when presenting features similar to activated B-lymphocytes. This subdivision is relevant for therapy and prognosis, as ABC cases show a worse outcome as regards progression-free survival (PFS) and OS after treatment with R-CHOP standard therapy \[[@B14-cancers-12-00185],[@B15-cancers-12-00185],[@B16-cancers-12-00185]\] in comparison to GCB patients.
However, GEP through microarrays poses a challenge because it is available only for a small fraction of patients whose mRNA can be extracted from fresh or frozen tissues. The attempts to substitute GEP with immunohistochemistry (IHC) applicable to formalin-fixed, paraffin-embedded (FFPE) tissue samples \[[@B17-cancers-12-00185],[@B18-cancers-12-00185],[@B19-cancers-12-00185],[@B20-cancers-12-00185],[@B21-cancers-12-00185],[@B22-cancers-12-00185]\] evidenced another series of inherent difficulties linked to the extreme variability of results, even when the same algorithm (Hans, Choi, Colomo, Muris, Pileri, or Tally) was applied \[[@B23-cancers-12-00185]\]. Indeed, when the two techniques were compared, it was evident that the classification of DLBCL based on the COO was different.
Five years ago a new approach named Lymph2Cx was proposed for GCB/ABC COO classification; based on a panel of 20 genes and applicable to mRNA extracted from FFPE tissue samples, it is conducted on the NanoString platform and replicates the results of conventional GEP, demonstrating its superiority to IHC algorithms in various series of DLBCL cases \[[@B24-cancers-12-00185],[@B25-cancers-12-00185],[@B26-cancers-12-00185],[@B27-cancers-12-00185]\].
In the context of GEP, a German research group recently analysed data generated from expression microarray analyses of 873 different types of lymphoma, including DLBCL, with the purpose of clarifying their phenotypic characteristics \[[@B28-cancers-12-00185]\]. Through this approach, the investigators demonstrated that the transcriptome panorama of B-cell lymphomas consists more in a continuum of expression states than of clearly separated phenotypes, and in which every layer represents a different lymphoma and individual cases \[[@B28-cancers-12-00185]\]. Focusing on DLBCL, a series of models employed GEP for classifying patients: Monti et al. subdivided cases into three groups that were identified in oxidative phosphorylation, B-cell receptor/proliferation, and host response \[[@B29-cancers-12-00185]\]; Dybkaer et al. isolated B cells from reactive tonsils and identified B-cell-associated gene signatures (BAGS), highlighting for each an association with molecular findings \[[@B30-cancers-12-00185]\]. Other researchers instead focused their efforts applying GEP in the study of the microenvironment of lymphoma cells \[[@B31-cancers-12-00185],[@B32-cancers-12-00185]\] and in the search of specific immune signatures \[[@B33-cancers-12-00185]\]. Indeed, over time a series of NanoString-based expression assays has been suggested; for example, the Michaelsen group recently proposed a new NanoString-based assay for BAGS classification to overcome the difficulty of microarray-based GEP, in order to better categorize DLBCL in diverse B-cell subtypes \[[@B34-cancers-12-00185]\]. Analysing microarray data from 970 patients belonging to four different cohorts, they first selected genes and then created and tested a new NanoString-based BAGS2Clinic assay for quick and easy-to-use BAGS profiling. Then, they tested the assay in an independent cohort of 88 lymph node biopsies and confirmed that it showed a good correspondence with the original BAGS classifier, having an overall accuracy of 84% and a subtype-specific accuracy ranging between 80% and 99%. It seems that BAGS classification could highlight important features of tumour biology and aspects about resistance to immuno- and chemotherapy that can be employed when choosing novel treatment strategies for DLBCL patients \[[@B34-cancers-12-00185]\].
However, albeit meticulous, even COO classification through GEP reserves diverse exceptions, and variation in patient outcome persists even within each COO subtype, posing the main difficulty in management of patients and raising the question of the deep study of the extreme molecular heterogeneity that features DLBCL \[[@B35-cancers-12-00185]\]. Consequently, several studies in the last years attempted to discover further stratification parameters serving to make a better patient categorization, and additional to the existing ones that presented some critical issues. Many prognostic markers have been considered, such as MYC gene alterations, which characterize from 5% to 15% of de novo DLBCL and confer a worse prognosis and higher risk of central nervous system involvement \[[@B36-cancers-12-00185]\], as well as TP53 mutation, Epstein--Barr virus infection, CD5 expression, CD30 expression, BCL2 rearrangement or expression, MHC class II expression, and others \[[@B37-cancers-12-00185],[@B38-cancers-12-00185],[@B39-cancers-12-00185],[@B40-cancers-12-00185],[@B41-cancers-12-00185],[@B42-cancers-12-00185],[@B43-cancers-12-00185],[@B44-cancers-12-00185],[@B45-cancers-12-00185],[@B46-cancers-12-00185],[@B47-cancers-12-00185],[@B48-cancers-12-00185]\], each presenting with contradicting data regarding their prognostic relevance. All things considered, at the current moment, genomic and transcriptomic profiling through next-generation sequencing (NGS) is surely the most powerful tool to investigate the molecular heterogeneity, as well as to find new potential biomarkers useful for diagnosis, risk determination, and treatment choice in DLBCL. However, in clinical practice, most investigations are conducted through the so-called "low-throughput techniques" (e.g., fluorescence in situ hybridization---FISH) and microarrays. Such approaches are acceptable if considering the old classification procedures that restricted the analyses to the search for a limited number of alterations, but now that fine molecular profiling of each individual case is possible, the implementation of NGS is becoming necessary also in the daily clinical workflow, maybe restricting the analyses to the search for specific molecular alterations.
1.2. Discovering New Prognostic Biomarkers and Models {#sec1dot2-cancers-12-00185}
-----------------------------------------------------
In the last years a series of discovery NGS researches was conducted trying to collect sequencing-derived information and to find focal alterations predicting prognosis, or in some cases, to elaborate models for DLBCL classification and prognostication ([Table 1](#cancers-12-00185-t001){ref-type="table"}).
In a large comprehensive exome and transcriptome sequencing of 1001 DLBCL cases \[[@B49-cancers-12-00185]\] Reddy and colleagues identified a set of 150 driver genes, most of which were then functionally characterized with an unbiased CRISPR screen of DLBCL cell lines to define oncogenes promoting cell growth. Through this information they drew up a prognostic model based on the presence of genetic alterations that was found to be better than current prognostic methods such as COO determination, IPI, and dual MYC and BCL2 expression. According to this model, genetic and/or expression aberrations of MYC defined the patient group with the worst prognosis; on the contrary, CD70 alterations in GCB-DLBCLs characterized the group with the better outcome \[[@B49-cancers-12-00185]\].
In a similar work by Schmitz and colleagues, NGS was adopted to unveil driver genes with recurrent alterations \[[@B50-cancers-12-00185]\]. They performed whole-exome sequencing (WES), RNA-seq, gene copy number analysis and targeted resequencing of 372 genes of 574 DLBCL cases, mostly pre-treated (96.5%), of which 51.4% were ABC and 20% unclassified (non-ABC, non-GCB). They developed a specific algorithm to identify genetic subtypes on the basis of the co-occurrence of genetic aberrations, and defined four prevalent subtypes that they termed MCD (co-occurrence of MYD88L265P and CD79B mutations), BN2 (BCL6 fusions and NOTCH2 mutations), N1 (NOTCH1 mutations), and EZB (EZH2 mutations and BCL2 translocations). Each subtype featured differences in gene-expression signatures, sensitivity to immunochemotherapy and outcome, BN2 and EZB subtypes being associated with favourable survival, and MCD and N1 subtypes to inferior outcomes. Indeed, MCD and BN2 subtypes seemed to depend on a chronic active BCR signalling pathway, opening an option for targeted therapeutic inhibition \[[@B50-cancers-12-00185]\].
Another similar 2018 research by Chapuy et al. analysed 304 samples from DLBCL patients to find recurrent mutations, low-frequency alterations, somatic copy number alterations, and structural variants \[[@B51-cancers-12-00185]\] identifying a series of genetic drivers that led to another new molecular classification. Through consensus clustering they characterized five DLBCL subsets, which included a novel group of ABC-DLBCL lymphomas of extrafollicular/marginal zone origin with low-risk and associated to NOTCH2 mutations (C1); a group of ABC-DLBCL with gains in BCL2 and/or mutations in MYD88L265P, CD79B, PIM1, and PRDM1, and associated with an unfavourable outcome (C5); two distinct subsets of GCB-DLBCL with different outcomes and targetable alterations (C3 group with aberrations affecting PTEN and epigenetic mediators such as KMT2D, CREBBP, and EZH2 and poor outcomes; and C4 group with alterations of signal transducers such as BCR--PI3K, NF-κB, or RAS--JAK, of transcription activators such as BRAF and STAT3, in histone genes, and genes involved in immune evasion (CD83, CD70, and CD58), and with favourable outcomes); and an ABC/GCB-independent group with biallelic inactivation of TP53, CDKN2A loss, and genomic instability, associated with poor outcomes (C2). The characteristics of the five subgroups identified also correlate with outcome in an independent manner in regard to IPI, suggesting new chances of therapeutic options and providing a roadmap for the identification of actionable DLBCLs \[[@B51-cancers-12-00185]\].
Comparing the last two proposed molecular classifications it seems that the C1, C3, and C5 groups overlap with the BN2, EZB, and MCD groups of the work by Schmitz et al \[[@B50-cancers-12-00185]\]. However, there are also differences, such as the C2 and C4 groups that did not present similarities with those in the other research. Indeed, both classifications differed from the molecular classification by Reddy and colleagues, in which MYC status was correlated with clinical outcome \[[@B49-cancers-12-00185]\]. Indeed, each model presents some critical issues: the model by Reddy et al. gives a prognostic value to each individual marker assessed, but, given the lack of clear clusters, it is hard to use for therapeutic purposes. The drawback of the study by Schmitz is that is only focuses on ABC type; the study of Chapuy et al. instead encompasses all DLBCL, however conformational studies are still lacking.
The utility of investigating only MYD88/CD79B mutations to improve DLBCL classification and prognostication was explored in a set of 250 DLBCL cases \[[@B52-cancers-12-00185]\]. The authors analyzed MYD88/CD79B mutations through NGS or allele-specific PCR, MYC/BCL2/BCL6 rearrangements by FISH, and EBV infections by EBER-ISH, identifying MYD88 and CD79B mutations in 29.6% and 12.3%, MYC, BCL2, and BCL6 rearrangements in 10.6%, 13.6%, and 20.3%, and EBV in 11.7% of cases, respectively. This study revealed that MYD88-mutated cases presented a significantly inferior five-year OS compared to wild-type; indeed, patients without any of the analysed alterations showed a superior OS compared to others carrying at least one aberrancy. In multivariable analysis, evaluating clinical-pathologic characteristics, outcome, and prognosis according to IPI, MYD88 mutations retained the adverse prognostic impact. Thus, investigating MYD88 mutations in DLBCL presents clinical utility as they feature a distinct molecular subtype with adverse prognosis \[[@B52-cancers-12-00185]\].
Further, MYD88 mutations, together with CD79A, CD79B, and CARD11 aberrations, are known to trigger chronic activation of the B-cell receptor (BCR) signalling pathway \[[@B65-cancers-12-00185],[@B66-cancers-12-00185],[@B67-cancers-12-00185],[@B68-cancers-12-00185]\]. Immunoglobulin M (IgM) is expressed in 90% of the ABC-like DLBCL subtype, and together with CD79A and CD79B, constitutes the BCR signalling complex, another mechanism of BCR aberrant chronic activation \[[@B69-cancers-12-00185]\]. A recent DNA copy number analysis of 1000 DLBCL cases identified gains of 18q21.2 as the most frequent genetic alteration in the ABC-like group, and recognized the TCF4 (E2-2) transcription factor gene as the main target of the genomic aberrations \[[@B53-cancers-12-00185]\]. With in vitro and in vivo experiments the effects of TCF4 overexpression were studied, observing its binding to IGHM and MYC gene enhancers and the augmented expression of the corresponding transcripts and proteins. Indeed, it was demonstrated that inhibition of TCF4 activity through the BET inhibitor ARV771 triggered death of the ABC-like DLBCL cells. Thus, this information represents a rationale for the employment of BET inhibitors for the subset of patients carrying this alteration \[[@B53-cancers-12-00185]\].
All the above-mentioned studies allowed the use of genomic aberrations to identify subgroups associated with distinct clinical outcomes, but it is not always possible offer a genetic classification to all DLBCL patients. In the attempt to solve this problem a series of other researches have been carried out. A 2019 study by Wang et al. employed WES data to establish mutant-allele tumour heterogeneity (MATH) \[[@B57-cancers-12-00185]\]. Based on the median expression level, patients were divided into low and high MATH score classes in which the higher MATH score group was associated with a higher risk of progression as compared to a lower MATH score, both in the discovery and in the validation set. The authors conclude that MATH has a prognostic value that could be considered in the management of DLBCL patients; a higher score of MATH has proven to be an independent risk prognostic factor in predicting recurrence \[[@B57-cancers-12-00185]\].
Another recent work instead studied the somatic hypermutation (SHM) mutational activities showing that they delineated the COO in DLBCL \[[@B55-cancers-12-00185]\]. Normally SHM acts during B-cell development targeting an immunoglobulin variable region \[[@B70-cancers-12-00185]\]; altered SHM hits several of the DLBCL driver genes \[[@B70-cancers-12-00185],[@B71-cancers-12-00185],[@B72-cancers-12-00185]\]. Alkodsi et al. found that the expression of 36 SHM target genes featured four novel SHM subtypes, strongly associated and overlapping with genetic subtypes already characterized by Schmitz et al., and that were significantly associated with OS and PFS of DLBCL patients treated with immunochemotherapy \[[@B55-cancers-12-00185]\]. Their stratification separates the GCB-DLBCL class into two major subtypes: SHM1, characterized by a high frequency of not always concurrent BCL2 and MYC aberrations and mutations in chromatin modifying genes, and including cases with poor outcome after standard R-CHOP therapy, that could be directed to alternative therapies; and SHM3, featuring mutations in the JAK-STAT pathway and a better outcome to standard cure. The ABC class was divided into SHM4, presenting with BCL6 fusions and mutations in CD70 and BCL10; and SHM2, presenting with the worst outcome and characterized by mutations in the BCR signalling pathway, that could be treated with kinase inhibitors. Through multivariate analysis of survival, they revealed that the SHM subtypes conferred a prognostic impact independently from the COO classification and IPI. Moreover, a distinct clinical outcome was observed for the SHM subtypes in the same COO subtype, and interestingly, even within unclassified DLBCL. Furthermore, they identified associations of each SHM subtype with driver mutations and oncogenic signalling pathways, proposing the possibility of choosing targeted therapy. Thus, SHM pattern represents a marker for the molecular and clinical classification of DLBCL \[[@B55-cancers-12-00185]\].
Also Arthur et al. in 2018 focused on SHM activity and analysed through WGS a discovery cohort of 153 DLBCL tumour/normal pairs, and performed data validation on an internal validation cohort of 338 cases and on an external validation cohort of over 1000 additional cases to find frequently mutated coding and non-coding loci, likely targeted by aberrant SHM \[[@B56-cancers-12-00185],[@B73-cancers-12-00185]\]. Through further analysis of matched RNA sequencing (RNA-seq) data, they suggested the potential cis-regulatory effects on coding genes of the alterations found \[[@B71-cancers-12-00185],[@B74-cancers-12-00185]\]. These analyses revealed recurrent mutations in the 3'UTR of NFKBIZ, responsible for oncogene deregulation, and NF-κB pathway activation in the ABC subclass; instead, in the GCB subgroup they evidenced small amplifications accompanied by over-expression of FCGR2B (the Fcγ receptor protein IIB), associated with poor outcomes \[[@B56-cancers-12-00185]\].
1.3. Discovering Causes of Transformation and Chemoresistence {#sec1dot3-cancers-12-00185}
-------------------------------------------------------------
In 2014 Pasqualucci and researchers through WES and copy-number analysis performed a pioneering work highlighting aberrations of CDKN2A/B, MYC and TP53 as major drivers of transformation of follicular lymphoma (FL) to an aggressive malignancy, typically DLBCL \[[@B58-cancers-12-00185]\] ([Table 1](#cancers-12-00185-t001){ref-type="table"}). Further, a subsequent work showed that one third of transformed FL harbor a MYC rearrangement \[[@B75-cancers-12-00185]\]. More recently, Gonzalez-Rincon et al. performed targeted NGS on 22 matched samples of pre-transformed FL/transformed DLBCL patients and on 20 non-transformed FL patients \[[@B59-cancers-12-00185]\]. Through this approach they identified several recurrently mutated genes with roles in B-cell differentiation, GC architecture and migration that were enriched at transformation, such as LRP1B, GNA13, and, in particular, POU2AF1, whose mutations seemed to characterize transformed forms rather than de novo DLBCL cases. Overall, they observed that pre-transformed FLs samples were more mutated and presented greater subclonal heterogeneity than non-transformed forms. Specifically, four genes differed between patients who did and did not show transformation: NOTCH2, DTX1, UBE2A and HIST1H1E; the mutation of these genes was related to a shorter time to transformation. With this information, the authors conclude, it could be easier to identify patients at higher risk of transformation \[[@B59-cancers-12-00185]\].
With the purpose of investigating the pathogenic causes of chemoresistance and relapse in DLBCL, a 2014 work sequenced VDJ junctions in 14 pairs of matched diagnosis--relapse tumours \[[@B60-cancers-12-00185]\]. The results of this study proposed two mechanisms of clonal evolution in which the early-divergent mode found two distinct clones, the diagnostic one and the relapsing one, that diverged early; and the late-divergent mode, in which relapse clones descended directly from diagnostic clones with minor divergence. Indeed, they identified in epigenetic modifiers such as KMT2D the potential early driving mutation targets, and in immune escape alterations the relapse-associated events \[[@B60-cancers-12-00185]\]. A following research analyzed 38 R/R DLBCL biopsies obtained at the time of progression after immunochemotherapy with WES and compared the obtained mutation frequencies to an unrelated cohort of 138 diagnostic DLBCLs, with the aim of identifying relapse-associated genes. Through this approach they evidenced TP53, FOXO1, MLL3 (KMT2C), CCND3, NFKBIZ, and STAT6 as top candidate genes implicated in therapeutic resistance. Indeed, they detected mutations that may affect sensitivity to novel therapeutics, such as MYD88 and CD79B mutations in a portion of R/R ABC patients, and STAT6 mutations in one third of R/R GCB patients that were associated with activated JAK/STAT signaling, increased phospho-STAT6 protein expression, and increased expression of STAT6 target genes \[[@B61-cancers-12-00185]\]. Another research highlighted JAK-STAT pathway involvement in the relapsed samples; the authors performed WES on 14 matched primary/relapse samples from six DLBCL patients and recorded a mild increase of mutations in relapsed samples as compared to primary tumour specimens; 264 genes possibly related to therapy resistance were identified, such as tyrosine kinases, glycoproteins, and JAK-STAT pathway genes, as well as PIM1, SOCS1, and MYC, already known to be related to a risk for treatment failure \[[@B62-cancers-12-00185]\].
Furthermore, recently two other large-scale differential multi-omics studies were conducted on R/R DLBCL patients \[[@B63-cancers-12-00185]\]. In the former, Fornecher et al. integrated quantitative proteomics and targeted RNA-seq data obtained from a cohort of R/R versus chemosensitive DLBCL patients and listed a set of 22 transcripts/proteins pairs, whose expression levels significantly differed between the two groups. In this list appeared genes involved in metabolism such as Hexokinase 3, in the microenvironment such as IDO1, CXCL13, in cancer cells proliferation, migration and invasion or the BCR signalling pathway such as CD79B \[[@B63-cancers-12-00185]\]. In the latter research, Rushton et al. collected samples from 134 R/R patients enrolled in three clinical trials and performed a combination of exome sequencing and target panel sequencing of lymphoma-associated genes on circulating tumour DNA (ctDNA) extracted from plasma samples and tissue biopsies \[[@B64-cancers-12-00185]\]. They found that R/R patients were enriched for mutations in five genes; TP53, IL4R, HVCN1, RB1, and MS4A1. Apart from TP53, already described in R/R cases \[[@B54-cancers-12-00185]\], they focused on the others showing that IL4R mutations may trigger constitutively activation of JAK/STAT signalling and were associated to inferior OS; HVCN1 modulates the B-cell receptor (BCR) function, and truncated HVCN1 isoforms have been demonstrated to enhance BCR signaling; MS4A1 encodes CD20, the target of Rituximab, and its mutations either truncate CD20, or destabilize a common transmembrane helix. Collectively, they found that DLBCL patients with such mutations present a higher risk of treatment failure \[[@B64-cancers-12-00185]\].
1.4. Double Hit or Triple Hit B-Cell Lymphomas {#sec1dot4-cancers-12-00185}
----------------------------------------------
The debate about research into aggressive mature B-cell lymphomas with MYC, BCL2 and/or BLC6 aberrations, defined as high-grade B-cell lymphoma with double or triple hit (HGBL-DH/TH), deserves a special section as their detection constitutes a principal goal according to the last WHO classification \[[@B76-cancers-12-00185]\]. Quite recently, two research groups tried independently to recognize HGBL-DH/TH cases using gene expression signatures. Applying a gene expression--based classifier to a cohort of 928 DLBCL patients, Sha et al. identified a molecular high-grade (MHG) subgroup comprising 83 patients (9%), 75 of which were afferent to the GCB subtype \[[@B77-cancers-12-00185]\]. They revealed a subcategory of DH lymphomas and MYC rearrangement in one half of the total. GEP analysis identified proliferative features similar to centroblasts. PFS analysis at 36 months was 37% for the MHG subgroup after R-CHOP compared with 72% for others. Indeed, DH lymphomas not afferent to the MHG subgroup showed no evidence of a worse outcome than other GCB-like patients. Furthermore, they analysed the benefits of the addition of bortezomib to standard R-CHOP therapy; the collected data suggested a possible positive response to bortezomib \[[@B77-cancers-12-00185]\].
In another work, Ennishi et al. studied RNA-seq data deriving from 157 GCB-DLBCL cases, including 25 HGBL-DH/TH-BCL2 cases, to elaborate a gene expression signature identifying HGBL-DH/TH-BCL2 from other GCB-DLBCLs \[[@B78-cancers-12-00185]\]. Through this approach they elaborated a 104-gene panel with which 27% of all GCB-DLBCLs were grouped by the same expression signature, even if only one half harbored MYC and BCL2 rearrangements (HGBL-DH/TH-BCL2). They evidenced that, regardless of the HGBL-DH/TH-BCL2 status, the so-called double-hit signature-positive (DHITsig+) patients were characterized by inferior outcomes after immunochemotherapy as compared to negative patients. Indeed, they projected a new NanoString assay (DLBCL90) that should be useful in routine diagnostics to easily identify DHIT-positive cases \[[@B78-cancers-12-00185]\]. Later, Hilton et al evaluated through WES 20 DHITsig+GCB-DLBCL cases apparently lacking MYC and/or BCL2 rearrangements and revealed six tumours with cryptic MYC or BCL2 rearrangements that were FISH negative \[[@B79-cancers-12-00185]\]. Copy-number analysis revealed MYC and MIR17HG gains or amplifications, and focal deletions of the PVT1 promoter, both of which may contribute to dysregulation of MYC and its downstream pathways. These results support the role of the GEP signature for identifying GCB-DLBCL with poor outcomes \[[@B79-cancers-12-00185]\].
Through the expression signatures found, most HGBL-DH/TH were identified, emphasizing that, regardless of genetic or epigenetic aberrations, these patients present similar gene expression features. This observation highlights the concept that the mechanisms of alteration of the physiologic function of MYC and/or BCL2 are diverse and still emerging. Apart from structural aberrations such as translocation or gene amplification, the MYC role can be altered by transcriptional and post-transcriptional modifications, by the activation of enhancer/superenhancer elements \[[@B80-cancers-12-00185],[@B81-cancers-12-00185]\] as well as by mutations \[[@B82-cancers-12-00185]\]. So, it is well known that there are diverse MYC alteration mechanisms, and other new ones are being discovered. For example, very recently Gallardo et al. showed a novel mechanism of MYC signalling aberrant activation \[[@B83-cancers-12-00185]\]. They explored the biological consequences of overexpression of hnRNP K, an RNA-binding protein, whose expression is altered in cancer. They analysed the clinical implications of hnRNP K overexpression in 75 DLBCL patients without MYC alterations, observing hnRNP K overexpression in DLBCL patients even without MYC aberrations and its association with a short OS and PFS. Furthermore, hnRNP K overexpression in transgenic mice induced the development of lymphomas and reduced survival. Indeed, through global screening experiments and biochemical assays, they showed that hnRNP K is capable of post-transcriptionally and translationally regulating MYC. This aspect renders hnRNP K overexpressing-cells sensitive to BET-bromodomain-inhibition both in vitro and in transplantation models, opening out a new treatment strategy for DLBCL patients \[[@B83-cancers-12-00185]\].
From a diagnostic point of view, because sometimes HGBL-DH/TH DLBCL lacks aggressive morphological and/or immunohistochemical characteristics, the doubt arises whether FISH testing for MYC, BCL2, and BCL6 genes should be performed for every DLBCL case to detect DH status. A limitation to FISH testing was proposed only for GCB and double-protein expressor (DPE) (i.e., cases with MYC and BCL2 protein overexpression) DLBCL cases, thus reducing the analyses to 15% of patients \[[@B84-cancers-12-00185]\], but until now studies on large cohorts of patients are lacking. In their report, Scott et al. evaluated the prevalence of HGBL-DH/TH and the data resulting from FISH, COO (Lymph2Cx gene expression assay and/or Hans algorithm), and IHC testing in a large cohort of 1128 DLBCL cases deriving from three clinical trials and a population-based registry \[[@B85-cancers-12-00185]\]. Overall, 8% of the DLBCL analysed cases were HGBL-DH/TH and mostly GCB (13.3%) rather than ABC (1.7%). They demonstrated that the MYC rearrangement (MYC-R) featured 12.2% of cases that were mostly, but not totally, GCB DLBCLs: MYC-R alone and MYC/BCL6 HGBL-DH were observed in both ABC and GCB DLBCLs; instead, MYC/BCL2 and MYC/BCL2/BCL6 HGBL-DH/TH characterized only GCB. The data collected by the authors suggested that the best method for identifying all HGBL-DH/TH tumours is to perform FISH for the MYC rearrangement for all cases; when FISH testing is positive, BCL2 and BCL6 gene aberrations should be investigated \[[@B85-cancers-12-00185]\]. Another option is limiting FISH screening to GCB DLBCLs, thus reducing FISH experiments to half of DLBCL patients, still allowing the detection of about 99% HGBL-DH/TH with BCL2 rearrangements. However, this approach would prevent identifying rare MYC/BCL6 HGBL-DH \[[@B36-cancers-12-00185],[@B86-cancers-12-00185]\] and ABC/non-GCB cases with isolated MYC rearrangements. Indeed, the selection of DLBCL cases based on DPE status and/or COO did not allow about 35% of all HGBL-DH to be detected. Thus, FISH screening for MYC, BCL2, and BCL6 should be performed in routine diagnostics, together with gene expression assays and NGS; alternatively, the optimum is testing for MYC rearrangements, followed by BCL2 and BCL6 analyses if the former is positive \[[@B85-cancers-12-00185]\].
However, FISH testing for MYC rearrangements frequently does not allow breakpoint characterization and MYC-partner gene identification. In an interesting study conducted by Chong et al., targeted sequencing of MYC, BCL2, BCL6 and the immunoglobulin (IG) loci was applied in 112 DLBCL cases with a MYC aberration to explore the rearrangement at base pair resolution and to identify the partner gene identity \[[@B87-cancers-12-00185]\]. They characterized the partner gene in 88 cases and identified a breakpoint cluster region upstream of the MYC coding region and in intron 1. In this region, mostly breaks for translocations involving IGH (80%) occurred, whereas breaks involving non genic rearrangements were located downstream of the gene locus with different partners such as IGL and IGK. They identified BCL6, ZCCHC7, and RFTN1 as recurrent MYC partners, never previously described. Indeed, they tested two commercially available FISH break-apart assays for the search of MYC rearrangements, and found discordant data in 32% of the examined patients. In HGBL-DH cases most (65%) of the MYC rearrangements presented non-IG partners and the breakpoints were located outside the genic cluster region in 72% of cases. Furthermore, patients with de novo HGBL-DH and MYC-IG aberrations featured a trend toward progression and to shorter OS as compared to patients with MYC--non-IG rearrangements, thus associating MYC rearrangement architecture to the clinical outcome \[[@B87-cancers-12-00185]\]. More recently, another work confirmed these data: by analyzing a large cohort of 2383 DLBCL patients, Rosenwald et al. identified MYC-R in 264 (11%) cases, and evidenced that the negative prognostic impact of MYC-R is largely observed in patients with MYC DH/TH disease in which MYC is translocated to an IG partner \[[@B88-cancers-12-00185]\].
HGBL-DH/TH lymphomas are a specific subcategory according to the WHO classification, owing to their particular worse outcome \[[@B89-cancers-12-00185]\]. As regards DH lymphomas, MYC and BCL2 rearrangements frequently trigger the corresponding protein overexpression, characterizing a specific group called DPE lymphomas \[[@B90-cancers-12-00185],[@B91-cancers-12-00185]\], clinically featuring rapid progression and poor outcome. Recently, a study by Uchida et al. showed the positive effect of the BCL2 inhibitor venetoclax for the therapy of this subgroup of lymphomas \[[@B92-cancers-12-00185]\]. In vitro studies on DH and DPE lymphoma-derived cell lines revealed that the survival of neoplastic cells seems to depend on BCL2 activity rather than that of MCL1, a protein with a pro-survival function. In this context, they demonstrated that venetoclax interrupts the interaction between BCL2 and BIM, a pro-apoptotic protein, induces dephosphorylation of BCL2, and represses MCL1 protein expression. In primary lymphoma cell cultures, venetoclax was able to induce apoptosis even at low doses \[[@B92-cancers-12-00185]\], showing venetoclax as a promising strategy for the treatment of DH-DPE lymphomas. Nonetheless, the first clinical studies showed that as monotherapy it probably has no room, and even when combined with a trail inhibitor doesn't seem effective in relapsed DLBCL \[[@B93-cancers-12-00185]\]. However, further investigations are needed before coming to definitive conclusions.
1.5. NGS Application in Clinical Practice: Liquid Biopsy {#sec1dot5-cancers-12-00185}
--------------------------------------------------------
In the future liquid biopsy will likely be the tool that can render NGS investigations more feasible and realistic. Analyzing circulating cell-free DNA (cfDNA) containing DNA released by the tumour cells (ctDNA), liquid biopsy is a non-invasive investigation that, joined to NGS sensitivity and specificity, will probably revolutionize cancer diagnosis, prognosis, and treatment. ctDNA has been demonstrated to be as accurate as genotyping of the diagnostic biopsy to detect somatic mutations in DLBCL \[[@B94-cancers-12-00185]\]; indeed, ctDNA analysis has proven able to define tumour burden \[[@B95-cancers-12-00185],[@B96-cancers-12-00185]\] and identify prognostic and actionable biomarkers \[[@B95-cancers-12-00185],[@B96-cancers-12-00185],[@B97-cancers-12-00185],[@B98-cancers-12-00185],[@B99-cancers-12-00185],[@B100-cancers-12-00185]\]. Indeed, it reflects the real tumour genomic heterogeneity, as demonstrated by the observation of varies mutations maybe originating from different tumour-associated localizations. Further, through liquid biopsy the response to therapy and minimal residual disease can be monitored, as well as transformation or chemoresistance emerging by tracking genetic evolution through ctDNA analysis over time \[[@B94-cancers-12-00185],[@B96-cancers-12-00185]\]. A 2019 work conducted liquid biopsy through targeted-NGS on a set of 390 lymphoma- and cancer-relevant genes in 50 lymphoma patients in order to establish the mutation profiles of different lymphoma subtypes and evaluate the correlation between the cfDNA concentration and other clinical indexes such as serum LDH and IPI \[[@B101-cancers-12-00185]\]. The cfDNA concentration in the plasma was significantly correlated with the clinical indices in DLBCL; indeed, the differences between GCB-DLBCL, non-GCB-DLBCL and natural killer/T-cell lymphoma were evident, confirming that NGS-based cfDNA mutation profiling is capable of discriminating different lymphoma subtypes as well as performing COO classification \[[@B96-cancers-12-00185],[@B101-cancers-12-00185]\], thus helping to direct precision medicine actions \[[@B101-cancers-12-00185]\].
Recently, ctDNA level measurement has recently been integrated in a new risk assessment method called CIRI (Continuous Individualized Risk Index), that dynamically evaluates individual outcome probabilities employing risk predictors obtained over time, producing real-time risk assessments during the patient's disease course. CIRI for monitoring DLBCL patients considers a total of six risk factors, including the IPI, COO, interim imaging (iPET), along with ctDNA measurements prior to cycles one, two, and three of therapy, and has been demonstrated to improve outcome prediction compared to conventional risk models, thus enabling therapy selection in the perspective of personalized medicine \[[@B102-cancers-12-00185]\].
1.6. Discovering Personalized Treatment Approaches {#sec1dot6-cancers-12-00185}
--------------------------------------------------
In the last years, the clinical management of patients with malignant lymphoma has benefited from research on tumour genomics and biology, particularly in the context of monoclonal antibodies and small molecule inhibitors \[[@B53-cancers-12-00185],[@B103-cancers-12-00185],[@B104-cancers-12-00185],[@B105-cancers-12-00185],[@B106-cancers-12-00185]\], despite some disappointing results of phase II/III trials on some promising agents (e.g., obinutuzumab and bortezomib) \[[@B107-cancers-12-00185]\] for which the underlying mechanisms are not well understood. Moreover, a problem of the first line studies is that in relapsed setting drugs have a short duration of responses and no plateau. Hence, any benefit seen in relapse setting does not necessarily translate in a durable increase in cure/remission in first line. Indeed, the efficacy of some drugs could be reduced because of incorrect combination with chemotherapeutic agents or insufficient dosage, as well as because of tumour-specific peculiarities \[[@B108-cancers-12-00185]\].
In the future, NGS implementation in the routine clinical diagnostics will render tumour genetic profiling within everyone's reach, offering the chance of a choice of tailored treatment strategies. However, despite the enormous excitement about the hypothesis of using targeted agents for patient personalized treatment, there are a series of inherent difficulties to be surmounted. Randomized trials on the addition of targeted drugs (ibrutinib, everolimus, bortezomib, and lenalidomide) to standard chemotherapy did not demonstrate a clear advantage \[[@B109-cancers-12-00185],[@B110-cancers-12-00185],[@B111-cancers-12-00185],[@B112-cancers-12-00185]\], and in some circumstances the use of targeted agents alone has been hypothesized. In other cases, the opportunity of drug combinations has been discussed \[[@B113-cancers-12-00185]\], as a better response can be obtained than with each drug alone. Indeed, although nowadays a number of new potent drugs are available, it is equally true that there could be a wide list of drug combinations that could be employed for patient treatment. The opportunity of using drug combinations that attack important cancer-signalling pathways at the same time from multiple fronts raises the chance of therapeutic success, especially in the treatment of tumours such as DLBCL that present with complex genetic heterogeneity. Indeed, this approach can reduce treatment resistance, which can frequently be due to pathway redundancies, cancer cell heterogeneity, and disease evolution \[[@B114-cancers-12-00185],[@B115-cancers-12-00185]\]. A clear example that confirms this concept is R-CHOP, which has recently been demonstrated to be effective and curative thanks to low cross-resistance, rather than synergy among drugs \[[@B116-cancers-12-00185]\].
However, today only a narrow list of approved drug combinations is available; indeed, they mostly derive from empirical clinical experience rather than rational design. To meet this need, some computational models have been developed that integrate the tumour genomic signatures with pharmacological profiles of drugs. In 2018, Preuer et al. developed a deep neural network model, DeepSynergy, to elaborate drug combinations integrating the data deriving from gene expression analysis of 39 cancer cell lines with the chemical peculiarities of 38 anti-cancer drugs \[[@B117-cancers-12-00185]\]. However, some authors observed that this method uses large numbers of known synergistic drug combinations, frequently not providing a hypothesis of the potential mechanism of a specific drug combination synergy. Other approaches were then developed in order to identify the underlying molecular mechanisms of disease; one example is Combinatorial Drug Assembler (CDA) \[[@B118-cancers-12-00185]\], a pathway-based model elaborated to discover drug combinations targeting pathways that overlap with tumour-enriched signalling pathways using differentially expressed genes. Another model is TIMMA \[[@B119-cancers-12-00185]\], which identifies drugs targeting multiple driver pathways by elaborating and combining drug screening data and drug target interactions into a target inhibition network framework. However, the survival pathways to be targeted were identified based on empirical selection, not considering genomic data.
Recently, some investigators proposed a new computational system biology tool that they called DrugComboExplorer, which combines specific genomic characteristics of cancer types (i.e., signalling pathways, interactome and pharmacological data) with pharmacogenomic profiles of 5585 drugs and bioactive compounds from the NIH LINCS program (Library of Integrated Network-based Cellular Signatures) \[[@B120-cancers-12-00185]\]. Indeed, by adopting a data-driven strategy and by combining multi-omics data (DNA seq, gene copy number, DNA methylation, and RNA-seq data) of individual cancer patients, this tool unveils new regulatory signalling pathway mechanisms (i.e., driver signalling networks) and is able to perform large-scale drug combination prediction (15,593,320 available drug combinations). In vitro validation experiments on DLBCL and prostate cancer cell lines, in order to evaluate the reliability and the predictive power of their bioinformatics tool, confirmed its utility in identifying targetable cancer driver pathways and prioritizing potential drug combinations useful to attack them \[[@B120-cancers-12-00185]\].
Surely, this kind of approach is still in its infancy as the authors admit, but further investigations are ongoing to try to apply these models to specific cancer cases in clinical practice in order to identify personalized drug combinations and more efficient treatment plans for individual patients \[[@B120-cancers-12-00185]\]. Then, large scale collaborations should be scheduled integrating mulit-omics data, Bayesian trial design, and early shared endpoints based on, for example, CIRI or any interim guided models to test in vivo the reliability of these drug combination-predicting models ([Figure 1](#cancers-12-00185-f001){ref-type="fig"}).
2. Conclusions {#sec2-cancers-12-00185}
==============
The attentive observer has surely realized that there is currently a dichotomy between the potentialities deriving from the recent discoveries for DLBCL diagnosis, prognosis, and treatment, and patient management in real clinical life. Despite discoveries of new drugs, R-CHOP remains the standard treatment approach, and targeted therapy is considered mostly for R/R patients. This circumstance will probably persist for a while, but when the biological background becomes clearer, when genome-wide screening has become within everyone's reach, and when targeted drugs have demonstrated their real benefits, personalized medicine will become feasible also for DLBCL.
In this panorama, a great improvement in clinical management of patients will certainly derive from the synergy of data obtained from liquid biopsy, providing information about therapy options stitched onto the patient's specific disease. Probably, today we are still far from this goal, as standardizations and clinical trial designs are still needed to render molecularly driven approaches really achievable. In any case, the bases are there, allowing us to pursue the goal of realizing targeted therapy for DLBCL.
The authors would like to thank Mary Victoria Pragnell, B.A., for language revision of the manuscript. This work was supported by the 'Il sorriso di Antonio' Association (Corato, Italy) and Associazione Italiana contro le Leucemie (AIL)-BARI.
The authors declare no conflict of interest.
{#cancers-12-00185-f001}
cancers-12-00185-t001_Table 1
######
Summary of the most relevant recent studies about DLBCL molecular classification and prognostication.
Reference Kind of Study Main Molecular Findings Clinical Implications
--------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
New Prognostic Biomarkers and Models
Reddy et al., *Cell* 2017 \[[@B49-cancers-12-00185]\] Exome and transcriptome sequencing of 1001 DLBCL cases Identification of 150 driver genes set, definition of a prognostic model better than current ones -MYC mutations or aberrant expression: worst prognosis-CD70 alterations: better outcome
Schmitz et al., *NEJM* 2018 \[[@B50-cancers-12-00185]\] WES, RNA-seq, gene copy number analysis and targeted sequencing of 372 genes in 574 DLBCL cases Development of a specific algorithm identifying four genetic subtypes: -MCD (MYD88L265P and CD79B mutations),-BN2 (BCL6 fusions and NOTCH2 mutations),-N1 (NOTCH1 mutations),-EZB (EZH2 mutations and BCL2 translocations) -BN2 and EZB: favourable outcome-MCD and N1: inferior outcome-MCD and BN2 subtypes depend on BCR signalling pathway activation (targeted therapeutic option)
Chapuy et al., *Nat Med* 2018 \[[@B51-cancers-12-00185]\] WES and targeted sequencing on 304 DLBCL patients Identification of five DLBCL subsets:-C1 (NOTCH2 mutations)-C2 (TP53 and CDKN2A alterations, genomic instability)-C3 (PTEN, KMT2D, CREBBP, and EZH2 aberrations)-C4 (BCR--PI3K, NF-κB, or RAS--JAK pathway alterations, BRAF, STAT3, CD83, CD70, and CD58 mutations)-C5 (BCL2, MYD88L265P, CD79B, PIM1, and PRDM1 alterations) -C1: low-risk-C2: poor outcome-C3: poor outcome-C4: favourable outcome-C5: unfavourable outcome
Vermaat et al., *Haematologica* 2019 \[[@B52-cancers-12-00185]\] NGS, allele-specific PCR and FISH on 250 DLBCL cases Identification of:-MYD88 and CD79B mutations in 29.6% and 12.3%-MYC, BCL2, and BCL6 rearrangements in 10.6%, 13.6%, and 20.3%, respectively MYD88 mutations: adverse prognostic impact
Jain et al., *Sci. Transl. Med* 2019 \[[@B53-cancers-12-00185]\] DNA copy number analysis of 1000 DLBCL cases Identification of 18q21.2 gains as the most frequent genetic alteration in the ABC-like group, with involvement of TCF4 (E2-2) transcription factor gene The inhibition of TCF4 activity through BET inhibitors could be employed in the treatment of this patient subset
Intlekofer et al., *Blood Cancer* 2018 \[[@B54-cancers-12-00185]\] Targeted NGS on 198 DLBCL cases Identification of a median number of six genetic aberrations per case, with 97% of patients presenting at least one alteration and 54% of cases more than one (e.g., MYD88, CREBBP, CD79B, EZH2) -Less common aberrations (BRAF, CD274 (PD-L1), IDH2, and JAK1/2) could be employed as potential therapeutic targets-TP53 alterations: more frequently associated to lack of response to first-line chemotherapy and involved in R/R DLBCL
Alkodsi et al., *Leukemia* 2019 \[[@B55-cancers-12-00185]\] WGS, RNA-seq, and gene expression from literature DLBCL cohorts The expression of 36 SHM target genes identifies four SHM subtypes:-SHM1 (BCL2, MYC, and chromatin modifying genes aberrations)-SHM2 (BCR signalling pathway mutations)-SHM3 (JAK-STAT pathway mutations)-SHM4 (BCL6 fusions and mutations in CD70 and BCL10) -SHM1: poor outcome after standard R-CHOP therapy-SHM2: worst outcome, could be treated with kinase inhibitors-SHM3: better outcome to standard cure-SHM4: worst outcome, similar to SHM2
Arthur et al., *Nat. Commun.* 2018 \[[@B56-cancers-12-00185]\] Integrative analysis of whole genomes, exomes, and transcriptomes on thousands of DLBCL cases Identification of:-recurrent NFKBIZ 3' UTR mutations causing NF-κB pathway activation in the ABC subgroup-Small amplifications associated with over-expression of FCGR2B, in the GCB subgroup These results revealed new driver DLBCL mutations, improving diagnostic assays and offering new possibilities for the development of targeted therapeutics
Wang et al., *Carcinogenesis* 2019 \[[@B57-cancers-12-00185]\] WES on 22 early stage DLBCL and validation on 35 primary DLBCL cases Identification of two MATH score classes: low and high MATH score groups according to the median expression level -The higher MATH score group was associated with a higher risk of progression-The MATH score has a prognostic value that could be considered in the management of DLBCL patients
Causes of Transformation and Chemoresistance
Pasqualucci et al., *Cell Rep.* 2014 \[[@B58-cancers-12-00185]\] WES and SNP array analysis on 12 FL samples at diagnosis and on 39 transformed FL Identification of CDKN2A/B, MYC and TP53 as major drivers of transformation of FL to an aggressive malignancy, typically DLBCL The genomic profile of transformed FL shares similarities with de novo DLBCL-GCB but also displays unique gene mutations with diagnostic and therapeutic implications
González-Rincón et al., *PLoS One* 2019 \[[@B59-cancers-12-00185]\] Targeted NGS on 22 pre-transformed /transformed and on 20 non-transformed FL cases Transformed FL are characterized by several recurrently mutated genes with roles in B-cell differentiation, GC architecture and migration (LRP1B, GNA13 and POU2AF1) -Four genes differed between patients who did and did not show transformation (NOTCH2, DTX1, UBE2A and HIST1H1E)-the mutation of these genes was related to a higher risk of transformation
Jiang et al., *Genome Biol.* 2014 \[[@B60-cancers-12-00185]\] High-throughput sequencing of rearranged VDJ junctions in 14 pairs of matched diagnosis-relapse DLBCL Two proposed mechanisms of clonal evolution:-the early-divergent mode with two distinct clones (the diagnostic and the relapsing one) that early diverged;-the late-divergent mode, in which relapse clones descended directly from diagnostic clones with minor divergence Although DLBCL relapse may result from multiple tumour evolutionary mechanisms, each mechanism could provide rationale for therapies
Morin et al., *Clin. Cancer Res.* 2016 \[[@B61-cancers-12-00185]\] WES on 38 R/R DLBCL biopsies and on an unrelated cohort of 138 diagnostic DLBCLs Identification of TP53, FOXO1, MLL3 (KMT2C), CCND3, NFKBIZ, and STAT6 as top candidate genes implicated in therapeutic resistance Detection of mutations (MYD88 and CD79B) that may affect sensitivity to novel therapeutics
Nijland et al., *Cancers (Basel)*. 2018 \[[@B62-cancers-12-00185]\] WES on 14 matched primary/relapse samples from six DLBCL patients Identification of 264 genes possibly related to therapy resistance, including tyrosine kinases, transmembrane glycoproteins, and genes involved in the JAK-STAT pathway Identification of resistance-related genes such as PIM1, SOCS1, and MYC, that confer a risk for treatment failure
Fornecker et al., *Sci. Rep.* 2019 \[[@B63-cancers-12-00185]\] Integrated quantitative proteomics and targeted RNA-sequencing in 8 R/R DLBCL cases versus 12 chemosensitive DLBCL patients Identification of a set of 22 transcripts/proteins pairs, whose expression levels significantly differed between the two analysed groups Identification of new biomarkers related to chemoresistance, new potential drug targets: Hexokinase 3, IDO1, CXCL13, S100 proteins, CD79B
Rushton et al., *Hematol. Oncol.* 2019 \[[@B64-cancers-12-00185]\] WES and targeted NGS on plasma samples and tissue biopsies from 134 R/R patients R/R patients were enriched for mutations in five genes: TP53, IL4R, HVCN1, RB1 and MS4A1 DLBCL patients with mutations in these five genes present a higher risk of treatment failure
Abbreviations: WES, whole-exome sequencing; RNA-seq, transcriptome sequencing; NGS, next-generation sequencing; R/R, relapse or refractory; MATH, mutant-allele tumour heterogeneity; SNP, single nucleotide polymorphism; FL, follicular lymphoma; VDJ, Variable Diversity Joining.
| {
"pile_set_name": "PubMed Central"
} |
Source files and CAD files are available in these two repositories: For OpenSCAD: <https://github.com/felipe-m/oscad_filter_stage> For FreeCAD: <https://github.com/felipe-m/freecad_filter_stage>.
Introduction {#sec001}
============
Over the last years there have been a movement towards creating and sharing Open Source Hardware (OSH). This trend has been empowered by open source hardware projects such as the RepRap 3D printers \[[@pone.0225795.ref001]\] and the Arduino platform \[[@pone.0225795.ref002]\], which have made manufacturing and electronic technology accessible and affordable. As a result to this movement many efforts have been made to define open source hardware and set its best practices \[[@pone.0225795.ref003]\] \[[@pone.0225795.ref004]\] \[[@pone.0225795.ref005]\].
Inspired by this movement, an engineering research area has emerged to develop open source scientific hardware and laboratory equipment \[[@pone.0225795.ref006]\] \[[@pone.0225795.ref007]\] \[[@pone.0225795.ref008]\] \[[@pone.0225795.ref009]\] \[[@pone.0225795.ref010]\]. Open source scientific hardware not only allows a more affordable laboratory equipment, but also contributes to the development of Open Science by facilitating the replication and comparison of the scientific experiments. Moreover, it favors the enhancement of experiments by letting others to improve and customize the devices for different purposes.
As stated in the Open Source Hardware Statement of Principles \[[@pone.0225795.ref003]\], open source hardware is hardware for which the design is made publicly available so that anyone can study, modify, distribute, make, and sell the design or hardware based on that design.
Making OSH is not a matter of just providing an open source license for the hardware; in addition, design files, documentation and any source code should be available in the preferred format for making modifications to them and with an acceptable open license. Furthermore, it is encouraged that these files are made editable with free and open source software (FOSS) \[[@pone.0225795.ref004]\] \[[@pone.0225795.ref007]\] \[[@pone.0225795.ref011]\] \[[@pone.0225795.ref012]\].
Ideally, in order to maximize the ability of individuals to use and make the hardware, open source hardware should provide unrestricted content, and use readily-available components and materials, standard processes, open infrastructure and open-source design tools \[[@pone.0225795.ref003]\] \[[@pone.0225795.ref007]\].
The Open Source Hardware Definition \[[@pone.0225795.ref003]\] is based on the Open Source Definition for Open Source software (OSS) \[[@pone.0225795.ref013]\] and adapted to the realms of tangible things. However, there are some aspects of hardware that differ from software. First, unlike software, most of a hardware project will fall within the scope of patent law rather than copyright law \[[@pone.0225795.ref014]\]. Secondly, hardware designs demand a wider range of expertise because new areas of knowledge are involved, particularly when including mechanics, electronics and software \[[@pone.0225795.ref015]\] \[[@pone.0225795.ref016]\].
And lastly, the source in hardware is not as clearly defined as in software, as Bonvoisin et al. show from their study of several OSH products \[[@pone.0225795.ref005]\]. Their analysis reveals a wide range of interpretations of open source hardware, exposing that many projects lack enough documentation to replicate or modify the product.
This last issue unveils one of the main problems for OSH: the deficiency in documentation to consider it truly open \[[@pone.0225795.ref005]\] \[[@pone.0225795.ref015]\] \[[@pone.0225795.ref017]\]. Bonvoisin et al. \[[@pone.0225795.ref005]\] assert that while software source code unambiguously defines the software, i.e. the product; the source for tangible things is not so clearly defined. Consequently, unlike OSH, the openness of OSS is implicitly fulfilled just by providing the source code. Although this statement can be considered accurate; it could nevertheless be argued that proper documentation for OSS is also a need, especially for large projects \[[@pone.0225795.ref018]\] \[[@pone.0225795.ref019]\]. It is unpractical to try to understand large and complex OSS projects without adequate documentation. The 2017 Open Source GitHub Survey supports this idea by highlighting that incomplete documentation is the biggest concern for OSS \[[@pone.0225795.ref020]\].
In addition to the stated OSH requirements, Oberloier et al. \[[@pone.0225795.ref007]\] encourage parametric design of OSH for scientific equipment. Parametric design enables customization by providing the flexibility to alter the model dimensions for other experimental purposes. Parametric models can be created using script-based computer-aided design (CAD) tools such as OpenSCAD \[[@pone.0225795.ref021]\]. OpenSCAD is an open source software tool that may be considered the defacto standard for OSH parametric design \[[@pone.0225795.ref008]\] as it has been widely used to create parametric OSH for laboratory equipment \[[@pone.0225795.ref007]\] \[[@pone.0225795.ref008]\] \[[@pone.0225795.ref022]\] \[[@pone.0225795.ref023]\] \[[@pone.0225795.ref024]\] \[[@pone.0225795.ref025]\] \[[@pone.0225795.ref026]\] \[[@pone.0225795.ref027]\].
The benefits of using a script-based CAD tool such as OpenSCAD are twofold. First, it allows model customization by means of a parametric design. Secondly, it provides a source code for the mechanical design, making it more similar to software. Thereby, addressing the hardware absence of source code that Bonvoisin et al. pointed out \[[@pone.0225795.ref005]\]. Having a source code for the hardware may mitigate the difficulties of OSH projects in defining their true openness. Besides, it allows OSH projects to use software management tools such as control version, software documentation and collaborative tools. In a way, the mechanical design would be similar to the electronic design of digital circuits with hardware description languages; although in the electronic field there is less availability of FOSS design tools to cover the whole process \[[@pone.0225795.ref016]\] \[[@pone.0225795.ref028]\]
Unfortunately, one of the major limitations of OpenSCAD is its inability to export to industry standard CAD file formats such as STEP \[[@pone.0225795.ref029]\] \[[@pone.0225795.ref030]\]. Although OpenSCAD can export the models to tessellated formats that can be read in most CAD tools, these formats are only approximate because they have lost their parametric dimensions. As many authors suggest \[[@pone.0225795.ref012]\] \[[@pone.0225795.ref016]\] \[[@pone.0225795.ref031]\] \[[@pone.0225795.ref032]\], providing a standard file format is crucial since it allows others to modify the OSH design with different CAD tools. This is a critical issue since the user may not be familiar with a script-based CAD tool such as OpenSCAD.
In this paper we review script-based FOSS CAD tools in order to find an alternative to OpenSCAD. Among these tools, we have found FreeCAD \[[@pone.0225795.ref033]\] to be a suitable candidate, since it is able to export to standard parametric CAD formats. Although FreeCAD is mainly used through its graphical user interface (GUI), it also allows creating CAD models using Python programming language \[[@pone.0225795.ref034]\]. Therefore, in this paper we analyze OpenSCAD, which is the most extensively used tool for modeling open scientific equipment with a programming language, and compare it with Python scripts for FreeCAD (hereafter FreeCAD Python).
In order to compare these tools, we have created a configurable OSH model of a motorized optical filter stage. The filter stage has four components that have been modeled with both OpenSCAD and FreeCAD Python. Modeling these parts with both tools has allowed us to analyze their strengths and weakness.
As a result of our analysis we suggest that although FreeCAD Python has a larger learning curve, it has an extensive set of features that makes it more suitable and powerful for modeling open source labware.
This paper is organized as follows. In the next section, script-based FOSS CAD tools are reviewed. Next, the CAD models used as a test-bench are presented. Afterwards, the parametrization of the CAD models is described. The following section discusses the benefits and drawbacks of the two CAD tools analyzed: OpenSCAD and FreeCAD Python. Conclusions are drawn in the final section.
Script-based CAD tools for parametric modeling {#sec002}
==============================================
In this study we analyze CAD tools for open source scientific equipment. Computer software for solid modeling can be broadly classified into two types: parametric and free form mesh modelers. The former create exact and complex mathematical data structures and models. The latter generate a simple mesh of polygonal surfaces, also known as tessellated geometries. Having the purpose of designing mechanical pieces, whose dimensions must be accurate, parametric modelers are better fitted for the task. In addition to using a parametric modeler, parametric modeling is also recommended for OSH labware \[[@pone.0225795.ref007]\]. In parametric modeling, designers define the size, shape and position of geometric features and assembly components in term of parameters \[[@pone.0225795.ref035]\] \[[@pone.0225795.ref036]\] \[[@pone.0225795.ref037]\] \[[@pone.0225795.ref038]\]. Since scripting is particularly appropriate for parametric design \[[@pone.0225795.ref031]\] \[[@pone.0225795.ref039]\] \[[@pone.0225795.ref040]\], we have limited the analysis to CAD tools with scripting capabilities. Finally, following the best practices for OSH \[[@pone.0225795.ref004]\] \[[@pone.0225795.ref007]\] \[[@pone.0225795.ref011]\] \[[@pone.0225795.ref012]\] we have only selected free and open source software.
We had initially chosen OpenSCAD for being the most widely CAD tool used in parametric modeling for open source scientific equipment. However, we needed to find an alternative because OpenSCAD is not able to export to a standard parametric CAD file format such as STEP \[[@pone.0225795.ref030]\].
There are other available FOSS CAD tools for solid modeling using a programming language. Examples of these tools are BRL-CAD \[[@pone.0225795.ref041]\], CadQuery \[[@pone.0225795.ref042]\], PythonOCC \[[@pone.0225795.ref043]\], FreeCAD \[[@pone.0225795.ref033]\], ImplicitCad \[[@pone.0225795.ref044]\], OpenJSCAD \[[@pone.0225795.ref045]\] and Blender \[[@pone.0225795.ref046]\].
Blender was rejected for being a free form mesh modeler, which is more suitable to create natural or artistic designs rather than precise mechanical designs.
Both OpenJSCAD and ImplicitCad are similar to OpenSCAD. OpenJSCAD uses JavaScript as a programming language. ImplicitCad is an entirely separate project from OpenSCAD, but it has an OpenSCAD language interpreter. Thus, OpenSCAD designs can be imported to ImplicitCad. However, neither OpenJSCAD nor ImplicitCad are able to export to standard parametric file formats. Therefore, they have not been considered in the analysis as they do not provide a solution for the main OpenSCAD disadvantage.
FreeCAD is a very active project that is able to export to standard parametric CAD formats. Besides, it allows both GUI modeling and script-based modeling using Python. FreeCAD can be totally controlled by Python scripts and provides an Application Programming Interface (API) for solid modeling using Python scripts. Therefore, FreeCAD can be used by both graphical designers and CAD programmers.
PythonOCC is a Python library that provides 3D modeling features. PythonOCC is a wrapper of the OCCT library \[[@pone.0225795.ref047]\], which is the same geometric modeling kernel that FreeCAD uses. PythonOCC is able to export to standard parametric formats; nonetheless, as a Python library without graphical interface, it may be too challenging for designers with less programming experience.
BRL-CAD is a powerful solid modeler that has been active for more than 30 years; however, it is an expert oriented tool with a long learning curve. Since the OSH labware designer is not necessarily an expert CAD designer, we consider that an easier tool would be preferable.
CadQuery is a Python library that allows creating parametric models with a reduced amount of code. CadQuery has two working versions: v1.2 \[[@pone.0225795.ref042]\] and v2.0 \[[@pone.0225795.ref048]\]. The former is built on top of FreeCAD API and can be installed as a FreeCAD workbench. The latter is built on PythonOCC. Both versions are able to export to standard parametric formats.
CadQuery v1.2 can be easily integrated into FreeCAD and be used with FreeCAD graphical interface. Thus, CadQuery v1.2 will be included in the analysis as a part of FreeCAD. On the other hand, CadQuery v2.0 is a new fork of CadQuery that is independent from FreeCAD. We have not included it in our analysis because it has been recently released (December 2018) and it is in an early development stage; however, we consider that it could be a promising alternative. At the present day the two versions remain compatible; hence, the CadQuery filter stage model and the code snippets shown in this paper are valid for both versions.
[Table 1](#pone.0225795.t001){ref-type="table"} shows a summary of the FOSS CAD tools with scripting capabilities. For each tool, the central column shows if it is able to export to the STEP standard parametric format. In addition, the last column shows an important characteristic of the tool.
10.1371/journal.pone.0225795.t001
###### FOSS CAD tools with scripting capabilities.
{#pone.0225795.t001g}
CAD tool STEP export Main characteristic
--------------- ------------- ----------------------------------------
OpenSCAD No Widely used for parametric OSH labware
Blender No Not intended for mechanical CAD
OpenJSCAD No JavaScript, similar to OpenSCAD
ImplictiCad No Similar to OpenSCAD
FreeCAD Yes Both GUI and scripted modeling
PythonOCC Yes Just for scripting, no GUI
BRL-CAD Yes Expert oriented
CadQuery v1.2 Yes FreeCAD workbench
CadQuery v2.0 Yes Recently released, built on PythonOCC
Positive features for designing OSH labware are shaded in light blue. Negative features are shaded in light red. Non shaded cell describe neutral features or that could be considered positive or negative.
As it has been said, OpenSCAD is the prevalent CAD tool for OSH scientific equipment; however, OpenSCAD has a major drawback related to exporting to standard exchange formats. In order to find an alternative CAD tool, this article compares OpenSCAD with FreeCAD Python. In this analysis, we will also include CadQuery v1.2 as a part of FreeCAD workbenches.
Open hardware models created as test-bench {#sec003}
==========================================
In order to compare the proposed CAD scripting tools, we have designed a parametric OSH motorized optical filter stage. The stage allows positioning an optical filter along a linear axis. The stage has four printable pieces modeled in both OpenSCAD and FreeCAD Python. They have also been modeled using CadQuery library v1.2, which is installed as a workbench in FreeCAD. The source code of the models is available in two software repositories: one for the OpenSCAD models \[[@pone.0225795.ref049]\] and the other for the FreeCAD and CadQuery models \[[@pone.0225795.ref050]\].
These pieces are: a stepper motor bracket; an optical filter holder that can be attached to a linear guide; and two pieces that are parts of a timing belt tensioner. The models are configurable; hence, they can be easily modified by just changing the model parameters. [Fig 1](#pone.0225795.g001){ref-type="fig"} shows an arrangement example of the configurable filter stage with its printable parts highlighted.
{#pone.0225795.g001}
Following, the printable pieces used for the analysis will be explained with more detail. For more information about the designs, the software repositories contain step-by-step tutorials explaining the generation of the models \[[@pone.0225795.ref049]\] \[[@pone.0225795.ref050]\].
Motor bracket {#sec004}
-------------
The purpose of this piece is mounting the stepper motor to a structure. It has four holes aligned to the mounting holes of the motor, and it has two slots to provide flexibility in securing the bracket to a surface. Besides, the bracket is reinforced along the sides to be able to hold heavy motors ([Fig 2A](#pone.0225795.g002){ref-type="fig"}). [Fig 2B](#pone.0225795.g002){ref-type="fig"} illustrates how the motor is mounted on an aluminum profile.
{#pone.0225795.g002}
Filter holder {#sec005}
-------------
The filter holder is a single printable piece to place an optical filter. It has various holes to attach the piece to a linear guide. It has more bolt holes than needed to provide flexibility to use it with different linear guides or other structures. The piece has two timing belt clamps to pull the filter along the desired direction ([Fig 3A](#pone.0225795.g003){ref-type="fig"}). [Fig 3B](#pone.0225795.g003){ref-type="fig"} shows the exploded view of the filter holder mounted on a linear guide.
{#pone.0225795.g003}
Belt tensioner {#sec006}
--------------
The belt tensioner is a more complex unit because it is composed by two printable parts and some other elements such as an idler pulley, and a few bolts, nuts and washers. The printable parts are an idler pulley tensioner and a tensioner holder. [Fig 4](#pone.0225795.g004){ref-type="fig"} shows the belt tensioner with its parts from two different perspectives. The tensioner holder (blue) and the idler tensioner (orange) are the printable parts.
{#pone.0225795.g004}
To help understanding of the belt tensioner assembly, [Fig 5](#pone.0225795.g005){ref-type="fig"} shows an exploded view of its parts.
{#pone.0225795.g005}
The belt tensioner works by turning the leadscrew. The leadscrew nut cannot rotate because it is inserted inside the idler tensioner; thus, depending on the direction of the leadscrew rotation, the belt tensioner will retract ([Fig 6A](#pone.0225795.g006){ref-type="fig"}) or extend ([Fig 6B](#pone.0225795.g006){ref-type="fig"}) the idler tensioner (orange). As a consequence, this operation tightens or loosens the timing belt.
{#pone.0225795.g006}
3D printing the models {#sec007}
----------------------
These four pieces have been designed to be 3D printed without support structures. Printing without support is faster, minimizes waste material, produces better surface finishing, reduces post processing work and thus, decreases the probability of damaging the piece due to the post processing.
[Fig 7](#pone.0225795.g007){ref-type="fig"} shows the orientation to print the pieces without support structures.
{#pone.0225795.g007}
Parametric design {#sec008}
=================
One of the main advantage of script-based modeling is the ability to change parameters values and generate variations of the original model with little effort. Programmed modeling can be time consuming compared to traditional graphical design, but the effort pays off when many variations of the model are needed, or when not all the specifications have been set from the beginning; thus, the final dimensions may change.
We have defined several parameters for the proposed models. Following, the main parameters for the motor bracket, the filter holder and the belt tensioner will be described.
Motor bracket parameters {#sec009}
------------------------
Depending on the stepper motor size the resulting bracket will have different dimensions; hence, the main parameter is the standardized NEMA size of the motor. For example, the NEMA size will define the parameter `motorbolt_sep` shown in [Fig 8](#pone.0225795.g008){ref-type="fig"} and will also determine the minimum inner space for the motor. Other parameters such as walls thickness, the length and width of the slots can be modified. Some of the main parameters are shown in [Fig 8](#pone.0225795.g008){ref-type="fig"}.
{#pone.0225795.g008}
As an example, [Fig 9](#pone.0225795.g009){ref-type="fig"} shows the resulting brackets for two different motor sizes. Note that the length of the slots have also been modified.
{#pone.0225795.g009}
Filter holder parameters {#sec010}
------------------------
The filter holder can be configured to carry different filter sizes and to be attached to almost any bolt arrangement. For this purpose, the bolt hole positions and sizes are configurable. [Fig 10](#pone.0225795.g010){ref-type="fig"} shows the main filter holder parameters. There are other parameters that are less critical and are documented in the source code.
{#pone.0225795.g010}
As an example, [Fig 11](#pone.0225795.g011){ref-type="fig"} shows two filter holders that have been generated with different parameters.
{#pone.0225795.g011}
Belt tensioner parameters {#sec011}
-------------------------
Unlike the motor bracket and the filter holder, the belt tensioner is composed by various elements; consequently, some of the dimensions of these elements are interdependent. In these cases, parametric design plays a significant role because it allows to establish the dependencies among the components. Thus, relieving the user from considering these relationships because they are automatically taken into account by the parametric model.
Basically, the tensioner holder depends on the idler tensioner dimensions, and the idler tensioner depends on the idler pulley dimensions. In particular, the parameters that determine the dimensions of the idler tensioner are:
- Idler pulley size.
- Tensioner stroke.
- Thickness of the walls.
- Leadscrew metric.
The parameters that determine the dimensions of the tensioner holder are:
- Idler tensioner size (determined by the aforementioned parameters).
- Belt height.
- Profile size where it is going to be mounted and its bolts size.
- Thickness of the walls.
Following, these parameters are explained in detail.
### Idler pulley {#sec012}
Some of the dimensions of the printable pieces depend on the idler pulley size. Idler pulleys can be easily acquired; but also, they can be made using bearings and washers, or they can be 3D printed. [Fig 12](#pone.0225795.g012){ref-type="fig"} shows an example of an idler pulley made of a specific bearing and some washers.
{#pone.0225795.g012}
The idler pulley can be made with different component sizes. For example, [Fig 13](#pone.0225795.g013){ref-type="fig"} shows two possible configurations and their influence on the idler tensioner size. The figure shows that the width of the idler tensioner is determined by the pulley size. Also, the space for the pulley will vary depending on the pulley size.
{#pone.0225795.g013}
### Tensioner stroke {#sec013}
The tensioner stroke defines the length of the tensioner; that is to say, how much it can be extended or retracted. [Fig 14](#pone.0225795.g014){ref-type="fig"} shows three different tensioner stroke values and their influence on the tensioner length.
{#pone.0225795.g014}
### Wall thickness {#sec014}
The thickness of the walls has an effect on the overall height of the tensioner ([Fig 15](#pone.0225795.g015){ref-type="fig"}). The wall thickness has also a small influence on the length.
{#pone.0225795.g015}
### Leadscrew diameter {#sec015}
Depending on the leadscrew diameter, a different size for the nut hole will be needed. This will have a small effect on the idler tensioner length ([Fig 16](#pone.0225795.g016){ref-type="fig"}). Obviously, the diameter of the leadscrew hole will also change.
{#pone.0225795.g016}
### Idler tensioner size {#sec016}
All the previous parameters determine the idler tensioner dimensions. Since the idler tensioner is coupled inside the tensioner holder, the idler tensioner size impacts on the size and shape of the tensioner holder. [Fig 17](#pone.0225795.g017){ref-type="fig"} shows two idler tensioners having different parametric values. The corresponding tensioner holders also result in different shapes and sizes. Note how the wall thickness is the same parameter for both the holder and the tensioner.
{#pone.0225795.g017}
### Belt height {#sec017}
This parameter sets the position of the bottom of the belt along the height of the tensioner ([Fig 18A](#pone.0225795.g018){ref-type="fig"}). Changes in this parameter will generate taller or shorter tensioner holders as shown in [Fig 18B](#pone.0225795.g018){ref-type="fig"}.
{#pone.0225795.g018}
### Base width {#sec018}
The base width or profile size indicates the profile size where the belt tensioner will be mounted. [Fig 19](#pone.0225795.g019){ref-type="fig"} shows the resulting belt tensioners for three different profile sizes.
{#pone.0225795.g019}
### Other parameters {#sec019}
There are other parameters but they do not have a considerable influence over the size. Examples of such parameters are the radius of the fillets or the size of the profile bolts. They are documented in the source code \[[@pone.0225795.ref049]\] \[[@pone.0225795.ref050]\].
Discussion {#sec020}
==========
In this section we compare OpenSCAD and FreeCAD Python. The analysis will include CadQuery v1.2 as a part of FreeCAD since, as stated above, CadQuery v1.2 can be added as an external FreeCAD workbench. Throughout this section, we will use the terms CadQuery workbench or just CadQuery to refer to CadQuery v1.2.
We have organized the discussion in four main topics: the geometric modeling kernel, the usability, the programming language characteristics and the tool features.
Geometric modeling kernel {#sec021}
-------------------------
Both OpenSCAD and FreeCAD use an external library as a geometric modeling kernel.
OpenSCAD uses the CGAL constructive solid geometry library \[[@pone.0225795.ref051]\]. In constructive solid geometry (CSG) solid models are created by applying successive operations to a set of basic shapes. These shapes are called primitives and the operations can be rigid motions (such as translation and rotation) or boolean operations (union, intersection and difference) \[[@pone.0225795.ref052]\] \[[@pone.0225795.ref053]\].
FreeCAD uses the Open Cascade Technology (OCCT) libraries \[[@pone.0225795.ref047]\]. OCCT is based on boundary representation (B-rep), in which objects are represented by their topological boundaries. A solid model is defined by a set of surface elements that delimit the boundary between the interior and the exterior of the solid. The surface elements are often defined by parametric equations \[[@pone.0225795.ref052]\] \[[@pone.0225795.ref053]\].
CadQuery v1.2 is a Python library built on top of FreeCAD API; consequently, it ultimately also uses the OCCT kernel.
One of the most serious disadvantages of OpenSCAD is the use of CGAL because CGAL works with polygonal meshes rather than parametric models. On the other hand, this problem is not present in FreeCAD because OCCT is a parametric modeler.
Easy of use {#sec022}
-----------
The usability has been divided into two subsections: how difficult is to step up the tool to just start coding and how easy is to model with the given language.
### Tool setup {#sec023}
OpenSCAD is extraordinarily easy to set up. OpenSCAD has an editor window where any code can be tested; thus, it is a matter of start coding and then, clicking a button to save, preview, render or generate the resulting STL file of the model. The code can be saved in a text file with an \*.scad extension.
Moreover, OpenSCAD has some model examples that can be loaded from the menu. This is a simple but a very useful characteristic to test the tool capabilities and learn by example.
OpenSCAD allows to have file dependencies in order to keep commonly used functions or constants in a different file. To define the location of these files, the relative path has to be included in the appropriate command.
On the other hand, using Python scripts in FreeCAD is not that simple. Actually, the tutorial about using Python in FreeCAD is placed in a section called "Power users hub" \[[@pone.0225795.ref054]\], what suggest that FreeCAD is not intended to be used in this way by the novice.
Since modeling in FreeCAD is primarily conceived to be done graphically instead of programmed, there is no code editor window when starting FreeCAD as in OpenSCAD. Instead, we have found two options.
The first option is the Macro Editor, where Python scripts can be executed and saved in the *User macro* directory.
The second option is the Python Console, where any command can be executed; however, it is a console, not a file editor where the user can save their design in a text file, as it is in OpenSCAD. From this console, any Python file can be loaded, but it is not straightforward. Similarly, loading other files dependencies is not direct unless the files are kept inside some specific directories.
The associated FreeCAD Python software repository of the OSH filter stage design \[[@pone.0225795.ref050]\] includes further information about how to execute the scripts.
Alternatively, the CadQuery workbench can be easily installed in FreeCAD through the Addon Manager, which is available from the graphical user interface. This workbench includes an editor window that allows editing and executing Python code. Furthermore, this workbench has several CadQuery example designs that can be loaded to learn by example.
CadQuery workbench and its editor can be used to design using both CadQuery and FreeCAD APIs. Therefore, it is a good place to start modeling with FreeCAD Python scripts in any of these two options available. Nonetheless, we find that the error messages given through the CadQuery workbench are more obscure than those given through FreeCAD Python console. Thus, we find debugging with CadQuery more difficult than using the FreeCAD console.
### Modeling {#sec024}
OpenSCAD language is similar to C programming language. OpenSCAD has a few basic 2D and 3D primitives (such as circle, square, polygon, cube, sphere, polyhedron and cylinder) and some operations and transformations. With these basic primitives and operations almost any technical piece can be modeled. Nevertheless, there are some operations that we missed, such as filleting or chamfering.
Since OpenSCAD has a limited set of primitives and functions, it is relatively easy to learn. There are good step-by-step tutorials that provide all the information to become skillful within a short time \[[@pone.0225795.ref021]\] \[[@pone.0225795.ref055]\].
As an example, [Fig 20A](#pone.0225795.g020){ref-type="fig"} shows the OpenSCAD code to model a box (rectangular cuboid).
{#pone.0225795.g020}
FreeCAD uses Python; thus, there is no need to learn a new programming language for those who already know it. FreeCAD offers a Python Application Programming Interface (API) to its OCCT kernel \[[@pone.0225795.ref047]\]. This API allows creating and accessing OCCT geometric primitives and functions. Essentially, there are two different kind of objects in FreeCAD Python: OCCT shapes and FreeCAD objects. OCCT shapes are the underlying OCCT solid models. On the other hand, FreeCAD objects link the OCCT shapes to their graphical representation. Therefore, in order to display an OCCT shape in FreeCAD graphical interface, there should be a FreeCAD object associated to that OCCT shape.
[Fig 20B](#pone.0225795.g020){ref-type="fig"} shows a FreeCAD Python script where these two objects are created. The OCCT shape (`sbox`) is created in line 5. The FreeCAD object is created in line 6; thus, drawing the OCCT shape in the graphical interface. In addition to these objects, the code include lines 2 and 3 to import libraries, and line 4 to create a FreeCAD document to be able to save it. Once the FreeCAD document is saved, not only the source code can be shared but also the FreeCAD model, allowing it to be modified through the FreeCAD graphical interface.
Alternatively, FreeCAD API offers functions that automatically create FreeCAD objects with its underlying OCCT shape. These functions hide the OCCT shape from the user. [Fig 20C](#pone.0225795.g020){ref-type="fig"} shows an example where a FreeCAD object (`fbox`) is created in line 4. This FreeCAD object already links to its OCCT shape, which can be accessed through an attribute. Lines 6 to 8 are used to assign the box dimensional values.
Although there are some tutorials about FreeCAD Python scripting \[[@pone.0225795.ref054]\], they are not comprehensive and the documentation is neither complete nor well organized. However, since FreeCAD GUI commands are just Python scripts, and these scripts can be redirected to FreeCAD Python console, an alternative way to learn is to create models with FreeCAD GUI commands and observe their corresponding Python scripts in the console.
CadQuery v1.2 is a Python library on top of the FreeCAD API. CadQuery eases parametric feature-based modeling by providing methods to facilitate the location and creation of features. Nevertheless, although the design approach is intuitive for simple or symmetrical models, we have found that it may be confusing for intricate pieces. CadQuery has well organized documentation and tutorials, what improves the learning process \[[@pone.0225795.ref056]\].
CadQuery creates its own type of objects for solid modeling. Since CadQuery is a library on top of the FreeCAD API, CadQuery objects also link to the underlying OCCT shapes. In addition, if we want to represent the shapes in the FreeCAD GUI, a FreeCAD object has also to be created.
[Fig 20D](#pone.0225795.g020){ref-type="fig"} shows an example of a CadQuery script. A CadQuery object is created in line 4. This object includes the underlying OCCT shape. Line 5 creates the FreeCAD object that draws the shape in FreeCAD GUI, although the user may not realize that a FreeCAD object has been created. Line 4 of [Fig 20D](#pone.0225795.g020){ref-type="fig"} exposes a differentiated CadQuery characteristic: CadQuery builds the geometry from defined planes (`Workplane`).
As a summary, [Fig 20](#pone.0225795.g020){ref-type="fig"} shows different codes to model a rectangular cuboid. The three FreeCAD Python examples (B, C, D) contrast with OpenSCAD simplicity (A) and illustrate the initial difficult that the newcomer may face using FreeCAD scripts. He will not only need to learn how to describe a model using a programming language, but also he would have to understand the intricacies of the different objects, methods, libraries and the underlying OCCT kernel. It may not be a problem if the designer has a programming background, but if the designer has little experience in programming, the initial barrier could be high.
Programming language characteristics {#sec025}
------------------------------------
In this subsection, some of the most significant programming language characteristics of both tools are compared. The analysis includes the programming paradigm, the scope of variables, the data types and the libraries.
### Programming paradigm {#sec026}
Programming languages can be classified based on their features. OpenSCAD is a declarative, purely functional language \[[@pone.0225795.ref055]\]; whereas Python supports multiple programming paradigms, including procedural, object-oriented, and functional programming \[[@pone.0225795.ref034]\].
As a consequence of OpenSCAD functional programming paradigm, variables are set at compile time, not at run time. Therefore, variables keep a constant value during their entire lifetime. If a variable is assigned a value multiple times, only the last value is used in all places of the code. This characteristic can be confusing for programmers used to procedural languages such as C or Python. For example, in [Fig 21A](#pone.0225795.g021){ref-type="fig"} variable `x` will have a constant value of 10; consequently, this script will create two spheres of the same radius 10. The assignment in line 1 has no effect and may mislead the designer in thinking that the spheres will have different size. The sphere created in line 2 will have a radius of 10 even that the final assignment of variable `x` is made afterwards (line 3).
{#pone.0225795.g021}
For this same reason, the assignment `x = x + 1` is not valid in OpenSCAD ([Fig 21B](#pone.0225795.g021){ref-type="fig"}).
On the other hand, Python is a multi-paradigm language and variables can be modified at running time as in procedural languages. Furthermore, it can make use of other paradigms such as object oriented programming, which is extensively adopted in FreeCAD Python.
Functional languages have benefits such as being more predictable and less prone to bugs; however, the programmer used to procedural paradigms may find functional programming too rigid to make fully parametric designs.
### Scope of variables {#sec027}
Variables are created within a scope in OpenSCAD; thus, their values are not available outside that scope. [Fig 22A](#pone.0225795.g022){ref-type="fig"} shows a situation where the variable scope may produce a different behavior than expected. Since `y` is assigned both outside and inside the if statement, `y` will have two different values depending on the scope. As a consequence, the sphere in line 7 will not be affected by the assignment of `y` inside the if statement (line 4); thus, producing spheres of different size.
{#pone.0225795.g022}
Note how scripts in [Fig 22A and 22B](#pone.0225795.g022){ref-type="fig"} are equivalent. The differences lie in the assignments placement, but since OpenSCAD is a functional language, both descriptions generate the same model.
OpenSCAD scoping rules can make parametric models more complicated to design when establishing variable dependencies. For example, suppose we want to create a cylinder whose height depends on its radius value. If the radius is 5, the height will be 15; otherwise, the height will be 20. [Fig 23](#pone.0225795.g023){ref-type="fig"} shows shows three attempts to do this.
{#pone.0225795.g023}
The first attempt ([Fig 23A](#pone.0225795.g023){ref-type="fig"}) will not work because `h` has not been assigned outside the if statement; therefore, it will be undefined when calling the cylinder function.
The second script ([Fig 23B](#pone.0225795.g023){ref-type="fig"}) has solved the problem by creating the cylinder inside the if alternatives. Nevertheless, this solution lacks efficiency because the same function call has to be repeated. For this small example we had to repeat the cylinder function call in lines 4 and 7.
The third example ([Fig 23C](#pone.0225795.g023){ref-type="fig"}) uses the conditional `?` to avoid assigning `h` in an inner scope. Although we can resort to the conditional `?` to create variables in an outer scope, it can be cumbersome to use it when there are many alternatives or when there are more than one variable to assign. We have experienced this problem when defining the dependencies of the belt tensioner, as it can be seen in OpenSCAD file `kidler.scad` \[[@pone.0225795.ref049]\]. Nonetheless, it can be used efficiently in combination with vectors, as some libraries of technical components have done \[[@pone.0225795.ref057]\] \[[@pone.0225795.ref058]\] \[[@pone.0225795.ref059]\].
Python does not have this behavior. Variables can be updated anywhere in the code, and they keep their value within their scope. Unlike OpenSCAD, the programmer can define the scope of variables using local, global and nonlocal variables.
### Data types {#sec028}
OpenSCAD has a limited set of data types: number (64 bit floating point), boolean, string, range, vectors and undefined. There are not user defined types in OpenSCAD.
In contrast, Python has several data types. Python provides the standard built-in data types, but also other specialized data types defined in the Python standard library and other available modules. In addition, programmers can create their own data types.
The standard data types can be summarized in numbers (integer, floating point and complex), strings, lists, tuples, sets and dictionaries. Some of them can be very useful for managing information, like lists and dictionaries. We have extensively used them to define the dimensions of the components of our OSH test-bench.
For example, [Fig 24A](#pone.0225795.g024){ref-type="fig"} shows how a dictionary can be used to get the thickness of a DIN 125 washer. In [Fig 24B](#pone.0225795.g024){ref-type="fig"} a two-dimensional dictionary is used.
{#pone.0225795.g024}
It is possible to have relatively analogous structures in OpenSCAD to get same result. Like using the conditional `?` with vectors (similar to [Fig 23C](#pone.0225795.g023){ref-type="fig"}). However, Python data types allow much more flexibility; moreover, Python offers a set of optimized methods and functions to manipulate these data structures efficiently.
### Libraries {#sec029}
OpenSCAD has a limited set of functions; however, anyone can create libraries to facilitate the design process. OpenSCAD has some libraries available ranging from mathematical functions to the creation of useful shapes and mechanical parts \[[@pone.0225795.ref057]\] \[[@pone.0225795.ref058]\] \[[@pone.0225795.ref059]\] \[[@pone.0225795.ref060]\].
FreeCAD has libraries of components \[[@pone.0225795.ref057]\] \[[@pone.0225795.ref061]\] and also has modules to extend FreeCAD functionality, such as CadQuery \[[@pone.0225795.ref042]\]. These modules provide a wide set of tools varying from advanced modeling to the creation of all kind of mechanical objects. The modules are available through the graphical interface as workbenches \[[@pone.0225795.ref062]\] \[[@pone.0225795.ref063]\].
In addition to these specific FreeCAD extensions, the FreeCAD Python programmer can resort to the rich and versatile Python standard library and other specialized modules. Python libraries are developed by a vast community, much larger than the FreeCAD or OpenSCAD specific communities. This is of paramount importance because it provides innumerable standardized solutions to FreeCAD Python.
As an example, file manipulation in FreeCAD Python is made through Python libraries, not because FreeCAD developers implemented that functionality. Consequently, if a FreeCAD Python designer wants to include the ability to read data from a file, she would just need to use the appropriate Python library to read and parse that file. In contrast, adding the ability to read and manipulate files in OpenSCAD would require OpenSCAD developers to implement that functionality.
This characteristic allows Python programmers to count on libraries for innumerable tasks. Examples of this kind of tasks are reading text files to get parameter values; writing text files to generate reports or bill of materials; performing many kind of computations, such as mathematical, matrix and finite elements; error handling; working with data interchange formats, like JSON, XML, YAML; among many other tasks. As a result, Python libraries add a broad range of capabilities to FreeCAD Python.
Tool features {#sec030}
-------------
Both OpenSCAD and FreeCAD are multiplatform (Windows, MacOS and Linux) free and open source software. OpenSCAD is under GPL2 and FreeCAD under LGPL2+.
This subsection analyzes two important features: the graphical interface and the import/export capabilities. In addition, we explore some characteristics like the performance and the suitability for modeling complete systems. To end, we look into the status of the software project development.
### Graphical user interface (GUI) {#sec031}
OpenSCAD GUI has commands for visualization such as zoom, rotation, change of perspective, among a few other commands. There is the option to include axis and perform animations. OpenSCAD also allows changing the color and transparency of the objects, although it cannot be done through the GUI, but must be defined in the code. OpenSCAD also features a parameter customizer that allows changing the model parameters from the GUI.
On the other hand, since FreeCAD has been mainly devised to be used through its GUI, it has all kind of commands to visualize, measure and transform the models. Commands are organized in workbenches to perform related tasks. As an example, it has the Technical Draw Workbench, which produces basic technical drawings of the models similar to Figs [8](#pone.0225795.g008){ref-type="fig"} and [10](#pone.0225795.g010){ref-type="fig"}.
Another remarkable feature of FreeCAD is its ability to integrate the scripts as commands in the graphical interface, allowing designers to generate and customize the models through the GUI. Therefore, regular users would not need to work with the source code. As a conclusion, FreeCAD GUI is clearly more powerful and flexible than OpenSCAD.
### Import/Export capabilities {#sec032}
As we have said, OpenSCAD has limited export/import capabilities. OpenSCAD can import some 2D file formats but only imports 3D tessellated file formats such as STL. The problem with these file formats is that they no longer contain parametric information.
This is an important limitation when working in projects with different CAD tool users. The OpenSCAD designer cannot incorporate exact models from other CAD tools in her/his design.
Likewise, OpenSCAD can only export to tessellated 3D file formats. These formats are suitable for 3D printing, but not to get an exact representation of the model. From our point of view, this is one of the main OpenSCAD drawbacks, since it inhibits sharing exact dimensional models. It seems that OpenSCAD is more oriented to the creation of models for 3D printing, but not to create exact dimensional models of the pieces that can be used in other CAD tools.
Actually, as it has been said, the question is that OpenSCAD uses a kernel that works with polygonal meshes rather than parametric models. Therefore, despite that the design have been modeled parametrically in OpenSCAD, the parametric information is lost once the internal model is generated.
[Fig 25](#pone.0225795.g025){ref-type="fig"} visually explains this situation with an example. On the left we have the source codes for OpenSCAD ([Fig 25A](#pone.0225795.g025){ref-type="fig"}) and FreeCAD ([Fig 25B](#pone.0225795.g025){ref-type="fig"}). OpenSCAD can only generate 3D tessellated models, which can be exported to mesh file format such as a STL. As it can be observed, the tessellated model is made of a polygonal mesh, what implies that the parametric dimensions, such as the radius of the primitives (sphere, cone and cylinder), have been lost.
{#pone.0225795.g025}
On the other hand, FreeCAD generates a parametric model that can be exported to a standard parametric file format such as STEP. This file format preserves its original dimensional information and can be manipulated by most of CAD tools; that is to say, it can be used as a source for future modifications in other CAD tools. This is an important feature, since other potential users may not be interested in coding, or they use a CAD tool that cannot import the source code. In addition, FreeCAD can also generate the tessellated model used for production.
In addition to FreeCAD ability to import/export to standard parametric models, FreeCAD has a workbench to offer interoperability with OpenSCAD. This workbench contains functions to import and repair OpenSCAD models. Nevertheless, depending on the OpenSCAD primitives used, some of the parametric information may be lost.
### Speed {#sec033}
We have attempted to compare the time taken to generate the models in both tools. However, since OpenSCAD creates a tessellated model, the speed depends on the mesh resolution. The resolution can be defined by OpenSCAD special variables `$fa`, `$fs` and `$fn`. Variable `$fa` defines the minimum angle for a fragment, `$fs` defines the minimum size of a fragment.
On the other hand, there is no need to specify the mesh resolution in FreeCAD unless a tessellated model is needed. [Table 2](#pone.0225795.t002){ref-type="table"} shows the time in seconds that took FreeCAD to generate the parametric models. FreeCAD CadQuery models take a similar time than plain FreeCAD Python because they use the same FreeCAD API. Actually, some of CadQuery models used in this study took slightly less time because, in order to make the tutorials, the FreeCAD Python models include the graphical representation of every step of the construction process. Nevertheless, since the execution times are very similar, for clarity purposes, only the execution time of the plain FreeCAD Python models will be included.
10.1371/journal.pone.0225795.t002
###### Time to generate parametric models.
{#pone.0225795.t002g}
Parametric time OpenSCAD FreeCAD
------------------ --------------------------------------- ---------
Motor bracket [\*](#t002fn001){ref-type="table-fn"} 0.43 s
Filter holder [\*](#t002fn001){ref-type="table-fn"} 0.76 s
Idler tensioner [\*](#t002fn001){ref-type="table-fn"} 0.45 s
Tensioner holder [\*](#t002fn001){ref-type="table-fn"} 0.61 s
\* OpenSCAD does not generate parametric models.
In order to compare the speed, all the scripts have been executed in a computer with an Intel Core i7-4700Q CPU at 2.4GHz and 8 GB RAM running 64 bit Windows 10 Home. In FreeCAD Python the completion time has been measured with the function `datetime.now()` (code is available in \[[@pone.0225795.ref050]\]). In OpenSCAD the completion time has been obtained by the provided rendering report. This report does not provide fractions of seconds.
The tessellated models generated by both tools are different, as it can be appreciated in [Fig 25](#pone.0225795.g025){ref-type="fig"}. We have not found any option in OpenSCAD to generate a different kind of mesh. FreeCAD has three meshing options: standard, Mefisto and Netgen. Each option has their own parameters to adjust the resolution and sometimes the shape of the mesh. FreeCAD standard mesh seems to be the most similar to OpenSCAD mesh. FreeCAD standard mesh has two parameters to define the mesh resolution: the angular deviation (`ad`) measured in angular degrees, and the deviation of the surface (`sd`) measured in millimeters. The former limits the angle between subsequent segments in a polyline. The latter limits the distance between a curve and its tessellation \[[@pone.0225795.ref064]\]. Note that these parameters do not correspond to OpenSCAD special variables; although in both cases, the smaller the values are, the higher the resolution will be.
Starting from values that in our experience have provide a good 3D printing quality, we have increased the resolution of both meshes in order to have more data to compare. We have defined three levels of coarseness: coarse, normal and fine. [Table 3](#pone.0225795.t003){ref-type="table"} shows the chosen values for these parameters.
10.1371/journal.pone.0225795.t003
###### Selected parameter values to get different mesh resolutions.
{#pone.0225795.t003g}
OpenSCAD FreeCAD
-------- ---------- --------- ----- ---------
Coarse 6° 0.4 mm 30° 0.1 mm
Normal 6° 0.2 mm 15° 0.05 mm
Fine 3° 0.1 mm 3° 0.01 mm
[Table 4](#pone.0225795.t004){ref-type="table"} shows the number of elements (vertices, edges and facets) of each of the generated meshes of the motor bracket. From the table we can see that the meshes generated by OpenSCAD and FreeCAD are different, since for a similar number of vertices, FreeCAD meshes have a much larger number of edges and facets than OpenSCAD meshes.
10.1371/journal.pone.0225795.t004
###### Motor bracket: Number of vertices, edges and facets of the meshes and time to generate them.
{#pone.0225795.t004g}
Motor bracket OpenSCAD FreeCAD
------------------------------------------------------- ---------- --------- ------- -------- -------- --------
Vertices 479 807 1,568 402 722 3,394
Edges 719 1,211 2,352 1,242 2.202 10,218
Facets 242 406 786 828 1,468 6,812
**Total time** [^a^](#t004fn001){ref-type="table-fn"} 2 s 4 s 8 s 0.45 s 0.47 s 0.57 s
^a^FreeCAD time includes the generation of the parametric model ([Table 2](#pone.0225795.t002){ref-type="table"})
FreeCAD is clearly faster generating meshes, especially when the resolution is increased. OpenSCAD times rise at a much higher rate for finer resolutions. At the finest resolution FreeCAD is more than ten times faster.
FreeCAD generates the meshes from the parametric model. Therefore, once the parametric model has been generated, the parametric time ([Table 2](#pone.0225795.t002){ref-type="table"}) can be subtracted from the total time ([Table 4](#pone.0225795.t004){ref-type="table"}). Comparing the time values of these tables, it can be observed that FreeCAD only uses a fraction of the total time to generate the mesh of the motor bracket.
[Table 5](#pone.0225795.t005){ref-type="table"} shows the same information for the filter holder. The filter holder is the most complex piece. As a result, it has the largest amount of elements and it takes more time to generate the meshes. Again, it can be observed that for larger meshes FreeCAD is much faster than OpenSCAD. For the fine resolution is more than 50 times faster.
10.1371/journal.pone.0225795.t005
###### Filter holder: Number of vertices, edges and facets of the meshes and time to generate them.
{#pone.0225795.t005g}
Filter holder OpenSCAD FreeCAD
------------------------------------------------------- ---------- --------- ------- -------- -------- --------
Vertices 1,972 2,746 5,422 1,178 2,282 11,114
Edges 2,958 4,119 8,133 3,600 6,912 33,408
Facets 1,000 1,387 2,725 2,400 4.608 22,272
**Total time** [^a^](#t005fn001){ref-type="table-fn"} 30 s 45 s 102 s 0,87 s 0,90 s 1,47 s
^a^FreeCAD time includes the generation of the parametric model ([Table 2](#pone.0225795.t002){ref-type="table"})
The idler tensioner and the tensioner holder are not intricate pieces. Their complexity lies on their parameters dependencies, but once the pieces are created, the meshes are relatively simple. ([Table 6](#pone.0225795.t006){ref-type="table"}) shows their meshes size and the time to create them. The resulting data is in line with of these tables support the previous analysis.
10.1371/journal.pone.0225795.t006
###### Idler tensioner: Number of vertices, edges and facets of the meshes and time to generate them.
{#pone.0225795.t006g}
------------------------------------------------------- ---------- --------- -------- -------- -------- --------
Idler tensioner OpenSCAD FreeCAD
Coarse Normal Fine Coarse Normal Fine
Vertices 474 830 1,610 316 580 2,694
Edges 713 1,247 2,417 972 1,764 8,106
Facets 241 419 809 646 1,174 5,400
**Total time** [^a^](#t006fn001){ref-type="table-fn"} 4 s 7 s 13 s 0.54 s 0.54 s 0.61 s
Tensioner holder OpenSCAD FreeCAD
Coarse Normal Fine Coarse Normal Fine
Vertices 535 778 1,498 322 591 2,741
Edges 777 1,142 2,192 990 1,797 8,247
Facets 246 368 698 660 1,198 5,498
**Total time** [^a^](#t006fn001){ref-type="table-fn"} 6 s 8 s 16 s 0.63 s 0.64 s 0.71 s
------------------------------------------------------- ---------- --------- -------- -------- -------- --------
^a^FreeCAD time includes the generation of the parametric model ([Table 2](#pone.0225795.t002){ref-type="table"})
[Fig 26](#pone.0225795.g026){ref-type="fig"} summarizes the total time to generate the meshes. As can be seen in the graph, the worst case in FreeCAD (filter holder fine mesh) is slightly faster than the best case in OpenSCAD (motor bracket coarse mesh). Nevertheless, given that the meshes are not equivalent and the limited number of pieces used in the study, the results should be treated with caution.
{#pone.0225795.g026}
### System modeling {#sec034}
A complete mechanical system can be modeled with OpenSCAD and there are some very good examples in OpenSCAD gallery \[[@pone.0225795.ref065]\]. Nevertheless, we believe that OpenSCAD is more oriented to create solid models of individual pieces with the purpose of 3D printing. We find some arguments to support this idea.
First, a system model is useful to get a visual idea, document and share the complete device. However, OpenSCAD lacks a powerful graphical interface and no model transformations can be done through it.
Secondly, OpenSCAD import/export capabilities are limited to tessellated file formats; thus, OpenSCAD models cannot be integrated in a design of a different CAD tool without losing information. Conversely, a parametric model created in other CAD tool cannot be integrated in a OpenSCAD design without information loss.
An example of the relevance of the import/export capabilities for generating a system model is the collaboration between Electronic CAD (ECAD) and Mechanical CAD (MCAD) tools to generate models that integrate electrical and mechanical components \[[@pone.0225795.ref066]\]. FreeCAD parametric models of electronic components can be exported to electronic design tools such as KiCad \[[@pone.0225795.ref067]\]. As a result, KiCAD can generate the whole CAD model of the electronic board with its components. Finally, the resulting CAD model of the board could be integrated into FreeCAD in order to create the CAD model of the whole electromechanical system. There is even a FreeCAD workbench aiming to foster the collaboration between KiCad and FreeCAD \[[@pone.0225795.ref068]\]. Although there is a way to do it in OpenSCAD, it is not direct and the parametric information is lost.
Thirdly, OpenSCAD rendering speed slows in complex models, what supports the idea that it is not a tool to generate a whole system.
In contrast, FreeCAD is conceived to model both individual pieces or a complete system. It has a powerful GUI and it is mainly developed to be used in this way, but at the same time, it has powerful programming capabilities. The generated models can be exported and imported into standard formats, so CAD files can be incorporated and shared to the public without losing information. As an example, the complete parametric OSH system presented in this article ([Fig 1](#pone.0225795.g001){ref-type="fig"}) has been modeled using FreeCAD Python.
### Tool development {#sec035}
To end this subsection we analyze both projects in terms of their development. The objective is to compare the data publicly available in an attempt to obtain some metrics in order to characterize the development. We have chosen some of the metrics suggested by Crowston and Howison \[[@pone.0225795.ref069]\], such as the number of developers (contributors), the level of activity (commits), cycle time (releases) and popularity (downloads). We also include the number of lines of code to offer a intuitive metric of the project size. The information have been gathered from \[[@pone.0225795.ref070]\] \[[@pone.0225795.ref071]\] \[[@pone.0225795.ref072]\] and the projects website, although we could not find download statistics for OpenSCAD.
We observe from [Table 7](#pone.0225795.t007){ref-type="table"} that although both projects have a high level of activity, FreeCAD is notably more active and have frequents releases. That could be predicted since a graphical CAD tool would attract more users than a scripting CAD tool. However, from the analysis of the contributions on parametric OSH for scientific equipment, we would think that there probably are many more users of OpenSCAD than FreeCAD Python scripts.
10.1371/journal.pone.0225795.t007
###### General project metrics.
{#pone.0225795.t007g}
Metric OpenSCAD FreeCAD
--------------------------------------------------------------------------- ----------- --------------------
Lines of code[^a^](#t007fn001){ref-type="table-fn"} 129,641 3,181,812
Commits during last year[^b^](#t007fn002){ref-type="table-fn"} 1,327 3,469
Contributors during last year[^b^](#t007fn002){ref-type="table-fn"} 32 112
Latest release May 2019 (0.18.3) July 2019
Latest major release May 2019 (0.18.0) Mar. 2019
Previous major release Mar. 2015 (0.17.0) Apr. 2018
Downloads of latest major release[^c^](#t007fn003){ref-type="table-fn"} N/A (0.18) 1,298,693
Downloads of previous major release[^c^](#t007fn003){ref-type="table-fn"} N/A (0.17) 1,698,323
^a^Not including comment lines or blank lines.
^b^From Sep. 2018 to Sep. 2019.
^c^Including all minor releases. Downloads until Sep. 25th, 2019.
Summary {#sec036}
-------
To recapitulate, OpenSCAD is easy to setup and its modeling language hides the low level details related to the internal data structures and their interaction with the geometry kernel. Consequently, OpenSCAD is accessible to the non-expert programmer and allows designers to focus on the CAD modeling.
However, OpenSCAD functional programming paradigm and its scoping rules may seem cumbersome for those who are not used to them. Nonetheless, OpenSCAD main drawback is that its kernel works with tessellated models, in which the parametric geometry is lost. As a result, the interoperability with other CAD tools is dramatically reduced since it is not able to export to standard parametric formats.
On the other hand, FreeCAD Python users need a programming language background to deal with the setup and the intricate programming structures that are exposed. However, installing FreeCAD CadQuery workbench allows FreeCAD users to avoid the setup difficulties and provides them with a friendlier coding interface. Once this barrier is overcome, the FreeCAD Python designer can enjoy a design tool that provides a plethora of advantages, such as exporting to standard parametric formats, using CadQuery library and any other Python libraries, and incorporating the scripted models into its graphical user interface, among many others.
Finally, [Table 8](#pone.0225795.t008){ref-type="table"} outlines the comparison of both tools.
10.1371/journal.pone.0225795.t008
###### Tool characteristics summary.
The table is divided in the four topics: (1) geometric modeling kernel; (2) easy of use, (3) programming languages characteristics and (4) tool features.
{#pone.0225795.t008g}
Topic OpenSCAD FreeCAD Python
-------------------------------------------------------- ---------------------------------------------------------- --------------------------------------------------
KERN Computational Geometry Algorithms Library (CGAL) Open Cascade Technology (OCCT)
Based on Constructive Solid geometry (CSG) Based on boundary representation (B-rep)
Polygonal mesh model Parametric model (B-rep)
EASY Usable, easy to setup Complex setup---Improved with CadQuery workbench
Easy to learn: reduced set of primitives and functions Complex to learn: many kind of functions and objects
Easy to learn: higher abstraction level Complex to learn: dealing with kernel structure
Step-by-step thorough tutorials Few scripting tutorials---Good tutorials for CadQuery
PROG Parametric design, code based, version control Parametric design, code based, version control
Specific programming language Widespread programming language (Python)
Declarative, purely functional paradigm Multi-paradigm, including procedural and object-oriented
Cumbersome scoping rules Scope of variables can be defined
Data types: number, boolean, string, range, vector Whole range of Python data types and user defined
Specific OpenSCAD libraries Specific FreeCAD libraries
No libraries other than from OpenSCAD Python libraries provide solutions to countless problems
Not good at handling input/output text files All kind of functions to read/write text files
No fillet or chamfer transformations Fillet and chamfer transformations
Minkowski and Hull transformations No Minkowski or Hull transformations
TOOL Multiplatform FOSS under GPL2 Multiplatform FOSS under LGPL2+
No modifications through GUI All kind of transformations through the GUI
No script integration in GUI Scripts can be integrated in the GUI
No import/export to standard parametric formats Import/export to standard parametric formats
Slow model generation Fast model generation
Difficult for system modeling Suitable for system modeling
Vibrant community creating OSH scientific equipment Almost no one creating open source labware
Active software development project Very active software development project
KERN, geometric modeling kernel; EASY, Easy of use; PROG, programming language characteristics; TOOL, tool features.
Positive features are shaded in light blue. Negative features are shaded in light red. Non shaded cell describe neutral features or that could be considered positive or negative. Negative FreeCAD features that improve when using the CadQuery workbench are shaded yellow.
Conclusions {#sec037}
===========
Designing open source scientific hardware using a programming language offers two clear benefits: it allows parametric design and provides a source code for the hardware. Parametric is design is a highly desirable characteristic for open source labware because it enables customization to suit different experiment purposes. On the other hand, providing a source code for the hardware helps to make the hardware truly open by mending the lack of enough documentation that OSH have.
Designing open source hardware using Python for FreeCAD has distinct advantages over OpenSCAD. We consider that the longer learning curve of Python for FreeCAD is largely compensated by three major benefits. First, the ability to export to standard parametric CAD formats. Secondly, the usage of a widespread programming language with an extensive standard library. Lastly, the ability to use and integrate the generated models and the scripts in the FreeCAD graphical interface; thus, allowing non-programmers designers to use and configure the models.
In the light of these clear benefits, we hope that our analysis and companion step-by-step tutorials will encourage the scientific community to adopt Python for FreeCAD for modeling parametric open source scientific equipment.
Future work will involve (1) the creation of modules to isolate the CAD designer from the lower level kernel structures; (2) the integration of the scripted models in the FreeCAD graphical interface to allow non-programmers to easily parametrize the designs; and (3) exploration of methodologies to aid parametric system modeling.
10.1371/journal.pone.0225795.r001
Decision Letter 0
Al-Ameri
Talib
Academic Editor
© 2019 Talib Al-Ameri
2019
Talib Al-Ameri
This 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.
12 Sep 2019
PONE-D-19-22759
Parametric CAD modeling for open source scientific hardware: comparing OpenSCAD and FreeCAD Python scripts
PLOS ONE
Dear Mr Machado,
Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE's publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.
To facilitate a timely publication, your revised manuscript should be uploaded as soon as possible.
We would appreciate receiving your revised manuscript by Oct 27 2019 11:59PM. When you are ready to submit your revision, log on to <https://www.editorialmanager.com/pone/> and select the \'Submissions Needing Revision\' folder to locate your manuscript file.
If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.
To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: <http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols>
Please include the following items when submitting your revised manuscript:
A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled \'Response to Reviewers\'.A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled \'Revised Manuscript with Track Changes\'.An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled \'Manuscript\'.
Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.
We look forward to receiving your revised manuscript.
Kind regards,
Talib Al-Ameri, Ph.D
Academic Editor
PLOS ONE
Journal Requirements:
1\. When submitting your revision, we need you to address these additional requirements.
Please ensure that your manuscript meets PLOS ONE\'s style requirements, including those for file naming. The PLOS ONE style templates can be found at
<http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf> and <http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf>
2\. We suggest you thoroughly copyedit your manuscript for language usage, spelling, and grammar. If you do not know anyone who can help you do this, you may wish to consider employing a professional scientific editing service.
Whilst you may use any professional scientific editing service of your choice, PLOS has partnered with both American Journal Experts (AJE) and Editage to provide discounted services to PLOS authors. Both organizations have experience helping authors meet PLOS guidelines and can provide language editing, translation, manuscript formatting, and figure formatting to ensure your manuscript meets our submission guidelines. To take advantage of our partnership with AJE, visit the AJE website (<http://learn.aje.com/plos/>) for a 15% discount off AJE services. To take advantage of our partnership with Editage, visit the Editage website ([www.editage.com](http://www.editage.com)) and enter referral code PLOSEDIT for a 15% discount off Editage services. If the PLOS editorial team finds any language issues in text that either AJE or Editage has edited, the service provider will re-edit the text for free.
Upon resubmission, please provide the following:
The name of the colleague or the details of the professional service that edited your manuscriptA copy of your manuscript showing your changes by either highlighting them or using track changes (uploaded as a \*supporting information\* file)A clean copy of the edited manuscript (uploaded as the new \*manuscript\* file)
3. We note that Figures in your submission contain copyrighted images. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For more information, see our copyright guidelines: <http://journals.plos.org/plosone/s/licenses-and-copyright>.
We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:
1. You may seek permission from the original copyright holder of Figure(s) to publish the content specifically under the CC BY 4.0 license.
We recommend that you contact the original copyright holder with the Content Permission Form (<http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf>) and the following text:
"I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (<http://creativecommons.org/licenses/by/4.0/>). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form."
Please upload the completed Content Permission Form or other proof of granted permissions as an \"Other\" file with your submission.
In the figure caption of the copyrighted figure, please include the following text: "Reprinted from \[ref\] under a CC BY license, with permission from \[name of publisher\], original copyright \[original copyright year\]."
2. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder's requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.
\[Note: HTML markup is below. Please do not edit.\]
Reviewers\' comments:
Reviewer\'s Responses to Questions
**Comments to the Author**
1\. Is the manuscript technically sound, and do the data support the conclusions?
The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.
Reviewer \#1: Yes
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
2\. Has the statistical analysis been performed appropriately and rigorously?
Reviewer \#1: Yes
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
3\. Have the authors made all data underlying the findings in their manuscript fully available?
The [PLOS Data policy](http://www.plosone.org/static/policies.action#sharing) requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data---e.g. participant privacy or use of data from a third party---those must be specified.
Reviewer \#1: Yes
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
4\. Is the manuscript presented in an intelligible fashion and written in standard English?
PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.
Reviewer \#1: Yes
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
5\. Review Comments to the Author
Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)
Reviewer \#1: The manuscript mainly presents an analysis study of OpenSCAD and compare it with FreeCAD Python scripts. The authors also created a parametric open source hardware design to compare these tools.
The manuscript is of interest to the PlosOne readership. The introduction provides sufficient background and include all relevant references. The research design is appropriate. The methods adequately described. The results clearly presented. The scientific style of presentation is high enough.
However, the introduction is too long, I would recommend authors to summaries the motivation.
(Line 73) The authors mentioned some other FOSS CAD tools available to create solid models using a programming language, such as FreeCAD, PythonOCC, CadQuery and BRL-CAD. Among these CAD tools, Among these CAD tools, we have found FreeCAD to be the most active, having around a major release each year,..etc. Could you please compare the interesting features of OpenSCAD and FreeCAD with FreeCAD, PythonOCC, CadQuery and BRL-CAD.
(line 262) \*.scad
The main references are WWW, we recommend the authors to cite research articles if applicable. The readership usually couldn't reach web sites for some cases and that happens for many reasons for example changing/removing the web address when a company acquired by another compony.
The flowing references are a web address; 2, 3, 4, 7, 12, 14, 21,22,30, 34, 3,36,37,38,39,40, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 56, 57, and 58.
Reference \[5\] missing the doi.
Reference \[9\] missing the doi.
Reference \[13\] missing the doi.
Reference \[15\] is a book but missing the doi.
Reference \[16\] please check the order of the authors; Luis Felipe Rosado Murillo then Matti Veikko Pietari Kauttu.
Reference \[32\] missing doi. [doi.org/10.1007/978-3-662-44468-9_53](http://doi.org/10.1007/978-3-662-44468-9_53).
Reference \[42\] missing doi. 10.1201/9781315119601.
Reviewer \#2: After reviewing the manuscript entitled "Parametric CAD modeling for open source scientific hardware: comparing OpenSCAD and FreeCAD Python scripts" (PONE-D-19-22759).
Here are my reviewing points:
The manuscript comparing OpenSCAD and FreeCAD. The work did provide some insight that can potentially guide the developers to improve the CAD modeling using open source scientific hardware.
1\. The comparison was comprehensive and the authors provided sufficient resources.
2\. The authors provided sufficient and good literature background.
3\. The featured of both OpenSCAD and FreeCAD were analyzed fairly.
4\. The major drawback is that this manuscript fits to be a case study rather than a research article. However, in both cases it is still valuable and interesting.
5\. typos line (73) (The) should be (There).
6\. In table (7), the authors mentioned that it is divided into (three) topics, but it seems that there are (four) topics discussed in the table as mentioned in line (238), which are (the geometric modeling kernel, the usability, the programming language characteristics and the tool features), so its need to be corrected to (four) topics in the title of the table.
7\. The authors have to follow one format for the references, and I would suggest them to mentioned the access time for websites.
Reviewer \#3: I know several groups working on parametric CAD modeling for open source scientific hardware and they demonstrated progress however there is no single comprehensive study compares such type of CADs (in my knowledge).
The authors present an analysis study of prevalent modeling CADs (OpenSCAD and FreeCAD). They created a parametric open source hardware design to compare OpenSCAD and FreeCAD . The authers also describe methodology of comparison and they provided necessary scripts and supplementary documents.
The main advantage of this paper is the systematic analysis. It feels that the authors predicted readership's questions, so they put consequent answers. The paper is of interest for many PlosOne's auditory. The done works of other groups and researchers are accurately acknowledged, I would recommend to publish this manuscript after minor corrections.
My main concern is about the motivation for the authors to publish their study in a journal while they have already deposited scripts elsewhere (github). I would recommend authors to explain that in the introduction. The bibliography needs some attention according to the journal instructions.
\*\*\*\*\*\*\*\*\*\*
6\. PLOS authors have the option to publish the peer review history of their article ([what does this mean?](https://journals.plos.org/plosone/s/editorial-and-peer-review-process#loc-peer-review-history)). If published, this will include your full peer review and any attached files.
If you choose "no", your identity will remain anonymous but your review may still be made public.
**Do you want your identity to be public for this peer review?** For information about this choice, including consent withdrawal, please see our [Privacy Policy](https://www.plos.org/privacy-policy).
Reviewer \#1: No
Reviewer \#2: No
Reviewer \#3: No
\[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link \"View Attachments\". If this link does not appear, there are no attachment files to be viewed.\]
While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, <https://pacev2.apexcovantage.com/>. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at <figures@plos.org>. Please note that Supporting Information files do not need this step.
10.1371/journal.pone.0225795.r002
Author response to Decision Letter 0
28 Oct 2019
Reviewer \#1:
However, the introduction is too long, I would recommend authors to summaries the motivation.
\- We have attempted to reduce the introduction length, but since other reviewer believed that the introduction was adequate, we were afraid to remove important background information. Thus, we have barely been able to reduce it.
\-\--
(Line 73) The authors mentioned some other FOSS CAD tools available to create solid models using a programming language, such as FreeCAD, PythonOCC, CadQuery and BRL-CAD. Among these CAD tools, Among these CAD tools, we have found FreeCAD to be the most active, having around a major release each year,..etc. Could you please compare the interesting features of OpenSCAD and FreeCAD with FreeCAD, PythonOCC, CadQuery and BRL-CAD.
\- The reviewer\'s comment is relevant. We have created a new section with an extended explanation of the features of these FOSS CAD tools. In this section we have explained why we have chosen FreeCAD and also have included the FreeCAD CadQuery workbench in the subsequent analysis. This FreeCAD workbench facilitates the use of Python scripts in FreeCAD, and thus, supports the use of FreeCAD. We believe that with this addition the manuscript is more comprehensive. Besides, we have added included the CadQuery models in the software repository.
\-\--
(line 262) \*.scad
\- Corrected.
\-\--
The main references are WWW, we recommend the authors to cite research articles if applicable. The readership usually couldn't reach web sites for some cases and that happens for many reasons for example changing/removing the web address when a company acquired by another compony.
The flowing references are a web address; 2, 3, 4, 7, 12, 14, 21,22,30, 34, 3,36,37,38,39,40, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 56, 57, and 58.
\- The WWW are mainly from computer programs, associations and projects statistics that we have not been able to find research articles with that information. However, all the WWW references have been archived and should be available at any time with the information at the time of archive.
\-\--
Reference \[5\] missing the doi.
\- Corrected. In the revised manuscript, this reference is \[4\].
Reference \[9\] missing the doi.
\- Corrected. In the revised manuscript, this reference is \[8\].
Reference \[13\] missing the doi.
\- We could not find the doi, but we added where it is available. In the revised manuscript, this reference is \[12\].
Reference \[15\] is a book but missing the doi.
\- We could not find the doi, but the ISBN (9780133373905) could be added if necessary. In the revised manuscript, this reference is \[14\].
Reference \[16\] please check the order of the authors; Luis Felipe Rosado Murillo then Matti Veikko Pietari Kauttu.
\- Corrected. In the revised manuscript, this reference is \[15\].
Reference \[32\] missing doi. [doi.org/10.1007/978-3-662-44468-9_53](http://doi.org/10.1007/978-3-662-44468-9_53).
\- Corrected.
Reference \[42\] missing doi. 10.1201/9781315119601.
\- Corrected. In the revised manuscript, this reference is \[53\].
\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\--
Reviewer \#2:
5\. typos line (73) (The) should be (There).
\- Corrected.
6\. In table (7), the authors mentioned that it is divided into (three) topics, but it seems that there are (four) topics discussed in the table as mentioned in line (238), which are (the geometric modeling kernel, the usability, the programming language characteristics and the tool features), so its need to be corrected to (four) topics in the title of the table.
\- Corrected.
7\. The authors have to follow one format for the references, and I would suggest them to mentioned the access time for websites.
\- The format of the references have been corrected. We have added the archived link with its archive date in order to have a permanent link at the time of archive.
\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\--
Reviewer \#3:
My main concern is about the motivation for the authors to publish their study in a journal while they have already deposited scripts elsewhere (github). I would recommend authors to explain that in the introduction.
\- We have deposited the scripts in github to be in compliance with PLOS ONE software sharing policy: "We expect that all researchers submitting to PLOS submissions in which software is the central part of the manuscript will make all relevant software available without restrictions upon publication of the work".
In the third section (Open hardware models created as test-bench), in line 172 of the revised marked-up copy, we have included an explanation about the availability of the associated scripted models.
\-\--
The bibliography needs some attention according to the journal instructions.
\- Corrected.
######
Submitted filename: response_reviewers.pdf
######
Click here for additional data file.
10.1371/journal.pone.0225795.r003
Decision Letter 1
Al-Ameri
Talib
Academic Editor
© 2019 Talib Al-Ameri
2019
Talib Al-Ameri
This 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.
13 Nov 2019
Parametric CAD modeling for open source scientific hardware: comparing OpenSCAD and FreeCAD Python scripts
PONE-D-19-22759R1
Dear Dr. Machado,
We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.
Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.
Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at <https://www.editorialmanager.com/pone/>, click the \"Update My Information\" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at <authorbilling@plos.org>.
If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact <onepress@plos.org>.
With kind regards,
Talib Al-Ameri, Ph.D
Academic Editor
PLOS ONE
Additional Editor Comments (optional):
Reviewers\' comments:
Reviewer\'s Responses to Questions
**Comments to the Author**
1\. Reviewer 3 has repeated the same comments as in round 1. I think the authors have adequately addressed all comments raised in a previous round of review and I feel that this manuscript is now acceptable for publication.
Reviewer \#1: (No Response)
Reviewer \#2: All comments have been addressed
Reviewer \#3: All comments have been addressed
\*\*\*\*\*\*\*\*\*\*
2\. Is the manuscript technically sound, and do the data support the conclusions?
The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.
Reviewer \#1: (No Response)
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
3\. Has the statistical analysis been performed appropriately and rigorously?
Reviewer \#1: (No Response)
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
4\. Have the authors made all data underlying the findings in their manuscript fully available?
The [PLOS Data policy](http://www.plosone.org/static/policies.action#sharing) requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data---e.g. participant privacy or use of data from a third party---those must be specified.
Reviewer \#1: (No Response)
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
5\. Is the manuscript presented in an intelligible fashion and written in standard English?
PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.
Reviewer \#1: (No Response)
Reviewer \#2: Yes
Reviewer \#3: Yes
\*\*\*\*\*\*\*\*\*\*
6\. Review Comments to the Author
Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)
Reviewer \#1: (No Response)
Reviewer \#2: The authors have addressed my comments that I had raised in a previous round of review and I feel that this manuscript is now acceptable for publication.
The paper is well organized and clearly written. The technical quality of the manuscript is satisfactory. The length of the manuscript is now seems good, where most repeated sections are manipulated.
Reviewer \#3: (same comments of round 1)
\*\*\*\*\*\*\*\*\*\*
7\. PLOS authors have the option to publish the peer review history of their article ([what does this mean?](https://journals.plos.org/plosone/s/editorial-and-peer-review-process#loc-peer-review-history)). If published, this will include your full peer review and any attached files.
If you choose "no", your identity will remain anonymous but your review may still be made public.
**Do you want your identity to be public for this peer review?** For information about this choice, including consent withdrawal, please see our [Privacy Policy](https://www.plos.org/privacy-policy).
Reviewer \#1: No
Reviewer \#2: No
Reviewer \#3: Yes
10.1371/journal.pone.0225795.r004
Acceptance letter
Al-Ameri
Talib
Academic Editor
© 2019 Talib Al-Ameri
2019
Talib Al-Ameri
This 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.
19 Nov 2019
PONE-D-19-22759R1
Parametric CAD modeling for open source scientific hardware: comparing OpenSCAD and FreeCAD Python scripts
Dear Dr. Machado:
I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.
If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact <onepress@plos.org>.
For any other questions or concerns, please email <plosone@plos.org>.
Thank you for submitting your work to PLOS ONE.
With kind regards,
PLOS ONE Editorial Office Staff
on behalf of
Dr. Talib Al-Ameri
Academic Editor
PLOS ONE
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Inhaled corticosteroids (ICSs) are the standard treatment for asthmatic children. Non-adherence with prescribed ICS treatment clearly causes uncontrolled asthma ^[@ref-1]^. Feedback from parents is the traditional approach to assess the adherence to the treatment regimen; however, there is an overestimation by the patients of the remaining amount of medication ^[@ref-2]^. Even though integration of a dose counter into the inhaler device improves the tracking adherence to prescribed medication ^[@ref-3]^, the metered-dose inhalers (MDIs) currently available for ICS delivery do not offer integrated dose counters.
Weighing of the MDI canisters may be an alternative method to assess a patient\'s medication adherence. The present authors previously developed a linear regression equation using canister weight to calculate the number of doses actuated from the MDIs ^[@ref-4]^. Weighing of the canisters was implemented into our asthma care system in March 2013. This study was designed to assess a patient\'s medication adherence after the integration of regular weighing of the canisters into the routine service.
Methods
=======
Study design and subjects
-------------------------
A cohort study was carried out between May 2013 and April 2014. The inclusion criteria were children aged less than 8 years with a diagnosis of asthma who attended the Pediatric Allergy Clinic at Songklanagarind Hospital (Hat Yai, Songkhla, Thailand) and had exacerbation of asthma requiring hospitalization or an emergency department visit within the previous year. Patients who had a previous history of intubation or other chronic conditions were excluded.
The research protocol (REC 55-021-01-1-2) was approved by the Human Research Ethics Committee, Faculty of Medicine, Prince of Songkla University. Informed consent was obtained from the parents/guardians.
Asthma management
-----------------
Fluticasone propionate was selected as the ICS therapy for the study. A FLIXOTIDE® 125 Inhaler (GlaxoSmithKline) is a pressurized metered-dose inhaler, which delivers 125 microgram of fluticasone propionate per actuation. Each canister supplies 120 actuations. FLIXOTIDE® 125 Inhaler and BabyHALER® (GlaxoSmithKline), a device to help patients taking inhaled medicine, were prescribed to all participants. The dosage of fluticasone propionate was one actuation twice a day. The add-on asthma therapies were provided according to the GINA guideline ^[@ref-5]^. Participants needed to participate in a regular schedule every 8 weeks to obtain a new FLIXOTIDE® 125 inhaler. For patients who did not achieve adequate control or maintain the adherence rate, the inhalation technique and medication doses were revised at the time they visited the clinic. Exacerbation was defined as asthma deterioration that required treatment with systemic corticosteroids or emergency department utilization or hospitalization.
Adherence assessment
--------------------
The duration of adherence assessment was 24 weeks. Each participant received three canisters of FLIXOTIDE® 125 Inhaler. Parents were asked to collect the discarded inhalers at the 8-week (visit 1), 16-week (visit 2), and 24-week (visit 3) after recruitment. The discarded MDI canisters were weighed by a laboratory scale (Sartorius Basic®). The weight of each canister was replaced in the regression equation to calculate the number of doses actuated from the MDIs. A regression equation for a fluticasone propionate MDI canister gives the number (n) of doses actuated from the MDIs:
n = 276.16 -- (14.62 × canister weight). ^[@ref-4]^
Statistical analysis
--------------------
All of the statistical analyses were conducted with R software (version 3.3.2) by the R Foundation for Statistical Computing. Adherence in each 8-week interval was calculated as the amount of medication actuated divided by the amount prescribed. Percentage of adherence was reported as median and range.
Results
=======
A total of 52 asthmatic children participated in the study. The characteristics of the participants are shown in [Table 1](#T1){ref-type="table"}. Half of the participants were male. The median age was 52.7 months (range, 18.3--91.7) and the age at the onset of asthma was 12 months (range, 1.0--48.0). Parents were the major caregivers. In total, 32% of the participants had other allergic co-morbidities. Most of participants had received ICS therapy for longer than 3 months. At the end of 24 weeks, 44, 33, and 23 (total 100) discarded MDI canisters were collected from visits 1, 2, and 3, respectively. The remaining median weights of the discarded canisters from visits 1, 2, and 3 were 11.172g, 11.229g, and 11.113 g, respectively, and the median percentages of adherence were 96.8%, 96.3%, and 96.3%, respectively. In 11 discarded canisters (11%), the remaining medication was more than 30% of the labeled doses. Approximately 90% of the participants had no asthma exacerbation during 24-week study period ( [Table 2](#T2){ref-type="table"}).
###### Demographic data of asthmatic children.
-----------------------------------------------------------------
Variable Results (N=52)
------------------------------------------- ---------------------
Male, n (%) 29 (55.8)
Weight, kg, median (range) 18.2 (10.9--30.0)
Height, cm, median (range) 107.4 (84.6--127.5)
Age, months, median (range) 52.7 (18.3--91.7)
Age onset, months, median (range) 12 (1.0--48.0)
Allergic disease, n (%)\ 17 (32.7)\
• Atopic dermatitis\ 3 (5.8)\
• Allergic rhinitis\ 12 (23.1)\
• Food allergy 5 (9.6)
Caregiver, n (%)\ \
• Parents\ 44 (84.6)\
• Grandparents 7 (13.5)
Number of hospitalizations, times/\ 1 (1--5)
person, median (range)
Number of exacerbations within previous\ 3 (1--9)
year, times/person, median (range)
Duration from last exacerbation, months,\ 4 (3--11)
median (range)
Duration of ICS therapy, n (%)\ \
• 0--3 months\ 15 (28.9)\
• More than 3 months 37 (71.1)
Number of studied canisters, n (%)\ \
• Visit 1\ 44 (84.6)\
• Visit 2\ 33 (63.4)\
• Visit 3 23 (44.2)
-----------------------------------------------------------------
###### Canister weight and percentage of adherence.
------------------------------------------------------------------------------------------
Variables Visit 1 Visit 2 Visit 3
--------------------------------- ------------------ ------------------ ------------------
Number of canisters 44 33 23
New canister weight, g,\ 18.838\ 18.819\ 18.812\
median (range) (18.663--18.929) (18.607--18.907) (18.470--18.913)
Discard canister weight, g,\ 11.172\ 11.229\ 11.113\
median (range) (7.889--16.042) (7.825--17.146) (9.400--16.020)
Actuated doses calculated from\ 112.8\ 112.0\ 113.7\
canister weight, median (range) (41.6--160.8) (25.5--161.8) (42.0--138.7)
Percentage of adherence,\ 96.8%\ 96.3%\ 96.3%\
median (range) (33.0--143.6) (20.2--126.4) (36.8--118.7)
Number of participants with\ 1 2 1
exacerbation
------------------------------------------------------------------------------------------
######
Raw data of canister weight (original and discarded weight), results of the actuated dose equation and percent of adherence.
######
Click here for additional data file.
Copyright: © 2017 Chuenjit W et al.
2017
Data associated with the article are available under the terms of the Creative Commons Zero \"No rights reserved\" data waiver (CC0 1.0 Public domain dedication).
Discussion
==========
The present study demonstrated high adherence rates with low variations between the three visits. The percentage of discarded canisters, which had more than 30% of the labeled dosage of medication remaining, reduced from 22% in our previous cross-sectional study ^[@ref-6]^ to 11% in this study. Achieving good adherence in this cohort could be explained by the Hawthorn effect: the parents and participants knew that their adherence would be measured by weighing the canisters; therefore, the individuals modified or improved their adherence in response to the awareness of being observed.
Approximately 90% of the participants had no asthma exacerbation throughout the study period. In our previous study, only 59% of the patients had adequate control ^[@ref-6]^. Patients had access to the same educational and medication intervention, but the adherence rates were significantly different. Previous studies verified that the adherence rates had an association between lower adherence rates and poor asthma control ^[@ref-7],\ [@ref-8]^.
Although weighing of canisters was less accurate than a dose counter for measuring adherence ^[@ref-9]^, the present study demonstrated that a weight-remaining dose correlation could be used to determine the inhaler medication adherence in real life, and intensive monitoring of adherence was successful in achieving control.
In conclusion, our results demonstrated that high adherence rates were achieved after integration of canister weighing into the asthma care service. The present study highlighted the need to incorporate a method to monitor medical adherence in clinical practice, which may contribute to adequate asthma control.
Data availability
=================
The data referenced by this article are under copyright with the following copyright statement: Copyright: © 2017 Chuenjit W et al.
Data associated with the article are available under the terms of the Creative Commons Zero \"No rights reserved\" data waiver (CC0 1.0 Public domain dedication). <http://creativecommons.org/publicdomain/zero/1.0/>
**Dataset 1: Canister weight and percent of adherence.** Raw data of canister weight (original and discarded weight), results of the actuated dose equation and percent of adherence. doi, [10.5256/f1000research.10710.d150970](http://dx.doi.org/10.5256/f1000research.10710.d150970) ^[@ref-10]^
10.5256/f1000research.11547.r24513
Referee response for version 1
Piboonpocanun
Orathai
1
Referee
Division of Allergy and Immunology, Department of Pediatrics, Faculty of Medicine, Siriraj Hospital, Mahidol University, Bangkok, Thailand
**Competing interests:**No competing interests were disclosed.
9
8
2017
Version 1
The authors tried to use canister weight to calculate the number of doses actuated from the MDIs and to assess medical adherence to MDIs.
There are some issues which need clarification Please clarify why the numbers of collected MDI canisters were not the same in each visit.Please mentioned about patients who had exacerbation. Were they the same patients? Were these patients nonadherence to medication or did they have any specific problem?
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
10.5256/f1000research.11547.r22150
Referee response for version 1
Lee
Bee Wah
1
Referee
Department of Paediatrics, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore
**Competing interests:**No competing interests were disclosed.
24
4
2017
Version 1
This study evaluated adherence to ICS prophylaxis in asthmatic children by using weight of canister as a means to determine inhaler compliance, by calculating doses used by using a regression equation based on canister weight. The authors conclude that in the absence of an inbuilt dose counter, weighing the canister was a useful means of improving adherence to daily treatment with ICS.
Detailed comments: There were 52 children participating in the study. However, with each follow up visit (n=3)the number of canisters measure reduced progressively (44, 33, 23). Was there loss to follow up or did these patients fail to return their canisters? Could these patients be used as their comparative group in terms of asthma control over the period of follow up?For the subjects with the 11 canisters that had remaining medication, was their asthma control affected by non compliance?It may be more appropriate to use the term 'used' rather than 'discarded' canisters.Under the paragraph: adherence assessment. Some editing of sentences may improve readability: Parents were asked to 'return' the 'used inhalers at the 8 week.....
'The weight of the canister was replaced in the regression equation' to "The number of remaining doses
of remaining doses was extrapolated by using the weight of the used MDI canisters into the regression equation as shown:........
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
10.5256/f1000research.11547.r21967
Referee response for version 1
Manuyakorn
Wiparat
1
Referee
https://orcid.org/0000-0003-4278-1186
Division of Allergy, Immunology, and Rheumatology, Department of Pediatrics, Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok, Thailand
**Competing interests:**No competing interests were disclosed.
21
4
2017
Version 1
The article is well written. However, there is some minor comment.
1\. The authors mentioned that \"The metered-dose inhalers (MDIs) currently available for inhaled
corticosteroid delivery do not offer an integrated dose counter\". This statement was true at the time when this study was performed. But currently, there are some MDIs that have dose counter. So this statement needs some amendment.
2\. Method: the authors mentioned that \"Participants needed to participate in a regular schedule every 8 weeks to obtain a new FLIXOTIDE® 125 inhaler.\" And the duration of treatment was 24 weeks. So it means that each participant would receive 3 inhalers. A total of 52 children was participated. So I wonder why only 100 MDI canister was evaluated. It should be 52\*3=156?. This may need the clarification.
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
[^1]: All authors contributed to study design, interpretation of study findings, manuscript preparation, and approved the final manuscript. WC, VE and AY contributed to data acquisition and validation. PS was responsible for project management, data analysis and funding.
[^2]: **Competing interests:**No competing interests were disclosed.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Transcription factors (TFs) are central control points in gene regulatory networks. The DNA-binding domains of TFs can be divided into a number of structural families and recognize specific sequences of target genes.^[@R1]^ Zinc finger (ZF) proteins are one of the largest TF super-families in plants.^[@R2]^ Members of this super-family contain domains that are stabilized by the presence of one or more zinc ions, and several families of ZF domains have been identified, based on their structure and zinc coordination pattern.^[@R3]^
The DNA-binding protein with one finger (DOF) family is a plant-specific multigene family of TFs and since the discovery of the first member of the family in maize,^[@R4]^ numerous putative coding genes for DOF proteins have been detected in various species of plants ranging from lower plants such as the green alga *Chlamydomonas reinhardtii,* which has a single CrDOF-domain containing protein, and the moss *Physcomitrella patens* (19 PpDOF proteins)^[@R5]^ to the higher plants, including both angiosperms and gymnosperms such as potato (35 StDOF proteins),^[@R6]^ tomato (34 SlDOF proteins),^[@R7]^ barley (26 HvDOF proteins),^[@R8]^ and rice (30 OsDOF proteins).^[@R9]^ Phylogenetic analysis has shown that the number of DOF protein-coding genes is directly related to the complexity of the organism.^[@R8]^ In *Arabidopsis thaliana*, the 37 DOF coding genes comprise about 10% of all ZF-coding genes in that organism.^[@R10]^ The DOF protein family is mostly expanded during the evolution of the vascular plants through recurrent duplication of the ancestor genes after the divergence of green algae and the ancestors of terrestrial plants.^[@R5]^
Unlike many other ZF proteins, which contain several ZFs that cooperate in DNA and protein binding, DOF proteins exclusively harbor only one ZF. This domain is characterized by a highly conserved DNA-binding domain at the N-terminal region of the protein, consisting of 52 amino acids and defined by a C-x(2)-C-x(7)-\[CS\]-x(13)-C-x(2)-C-x-R-x-W-T-x-G-G motif (where x is any amino acid).^[@R11]^ Despite the similarity of the DOF-ZF consensus to ZFs in steroid hormone receptors and metazoan GATA ZFs ([Fig. 1](#F1){ref-type="fig"}), the DOF-ZF has a longer loop separating the putative zinc-coordinating cysteine pairs (the C-X~2~-C units). It has been shown that shortening of this loop in the DOF-ZF domain of ascorbate oxidase binding protein (AOBP) by deleting seven or more residues or replacement of the loop with that of zinc-binding unit 1 of estrogen receptor abolishes DNA binding activity.^[@R12]^ Substitution of the conserved cysteine residue near the center of the loop region of ZF domain of pumpkin AOBP improves DNA binding properties while substitution of the tryptophan residue close to the end of the same ZF domain ([Fig. 1](#F1){ref-type="fig"}) reduces DNA binding.^[@R13]^ So far, all reported DOF proteins for which DNA-binding data are available to recognize promoters containing cis-regulatory elements with the AAAG motif, except for pumpkin AOBP for which the recognition site is AAGT.^[@R11]^
{#F1}
DOF proteins are essential for the growth and development of plants and interestingly they mostly control biological processes that are exclusive to plants, such as vascular development, phytochrome signaling, seed germination, nitrogen assimilation, photosynthetic process and resistance to abiotic stresses.^[@R11]^ Although DOF proteins vary in sequence and function, their DOF-ZF domains show significant sequence similarity (Fig. S1). High similarity among the amino acid sequences of DOF-ZF domains suggests the similarity in their DNA binding specificity. However, although features of the interaction between DOF3.4-ZF from *A. thaliana* and its cognate DNA recognition sequence have been proposed based on molecular models built with GATA1 ZF proteins as templates, no experimental structural information is available either on DOF-ZF domains themselves or on their interaction with DNA.^[@R14]^
In the current study, a number of DOF-ZF domains from *A. thaliana*, namely DOF2.1, 3.4 and 5.8, were cloned, expressed, and purified in a bacterial expression system. One-dimensional NMR and fluorescence spectra confirmed that these recombinant domains are able to form stably folded structures and the binding ability of these domains to double-stranded oligonucleotides containing one and 2 copies of the consensus sequence AAAG was investigated. To the best of our knowledge, this is the first experimental report describing the quantitative measurement of the sequence-specific DNA binding affinity of DOF domains.
Materials and Methods {#s2}
=====================
Reagents {#s2-1}
--------
Qiagen RNeasy, and RevertAid cDNA synthesis kits were purchased from Qiagen (Hilden, Germany). Gel purification and plasmid mini extraction kits were obtained from Bioneer (South Korea). Primers were supplied from Bioron (Ludwigshafen, Germany). Glutathione Sepharose 4B was purchased from GE Healthcare Life Sciences (Daejeon, Sweden). 5'-Fluorescein-labelled and unlabelled oligonucleotides were purchased as synthetic oligonucleotides from Integrated DNA Technologies (NSW, Australia). Deuterium oxide (D~2~O, \>99.96% purity) was from Sigma (Missouri, USA) and 2, 2-dimethyl-2-silapentane-5-sulfonate sodium salt (DSS) was obtained from Fluka A.G. (Buchs, Switzerland).
Plant {#s2-2}
-----
*Arabidopsis thaliana* (ecotype: Columbia) seeds were obtained from the Faculty of Agriculture, Tabriz University, Iran. The seed preparation and growing were all carried out in a greenhouse at Tabriz University. The protocol for planting and growing *A. thaliana* was that used by the Arabidopsis Biological Resource Center with minor modifications (<http://abrc.osu.edu/seed-handling>).
Construction of the expression vector {#s2-3}
-------------------------------------
To construct the vectors used for DOF-ZF domain expression, total RNA from 3-week old *A. thaliana* was utilized. Nested PCR amplification was performed using 2 sets of primers ([Table 1](#T1){ref-type="table"}) designed based on sequence information for DOF2.1 (AT2G28510), DOF3.4 (At3g50410) and DOF5.8 (At5g66940) available at NCBI. Plasmids to express the DOF-ZF domains N-terminally fused to glutathione S-transferase were constructed by inserting *Bam*HI and *Eco*RI double digested PCR products between the same restriction sites in the pGEX-6p-1 vector. The accuracy of the constructs was verified by Sanger sequencing.
###### Oligonucleotide primers used for amplification of DOF ZF domains
------------------ --------- --------------------------------- -----------------------------------
**DOF proteins** **Outer primers** **Inner primers**
DOF2.1 Forward 5\'GTGCAGGAAATCTCAAACGAGAC3\' 5\'AAAGGATCCAGAGGAGAATTAGGAGG3\'
Reverse 5\'TGATGCTTTTGGAGAGTAGCGA3\' 5\'GAGGAGAATTCTCATCGACGGCAACCA3\'
DOF3.4 Forward 5\'CCAAATTCTCACTCTCTCATACCCT3\' 5\'GCGTAGGATCCCCGATTTCTGACC3\'
Reverse 5\'GAGGGAAGAGAACAGGCGTC3\' 5\'GAACGGAATTCTCATTTACGAGTACC3\'
DOF5.8 Forward 5\'ACGGCCAAGGAGGATCTGTTGC3\' 5\'TCTGGATCCATTCCGACGGATCAACAA3\'
Reverse 5\'CCGTCGTGATACCGCCGTTGG3\' 5\'GAACGGAATTCTCATTTACGGGAAAC3\'
------------------ --------- --------------------------------- -----------------------------------
Expression and purification of recombinant DOF-ZF proteins {#s2-4}
----------------------------------------------------------
Plasmids containing GST-DOF-ZF coding sequences were transformed and expressed in *Escherichia. coli* BL21 (DE3) cells. Bacteria were grown at 37°C in one liter of LB broth containing 100 µg/mL ampicillin to an optical density of 0.7 and expression of the GST-DOF protein induced by addition of 0.4 mM of isopropyl-1-thio-ß-D-galactopyranoside (IPTG) followed by overnight incubation at 20°C. Cells were harvested and resuspended in lysis buffer (50 mM Tris-HCl pH 7.2, 500 mM NaCl, 10% Triton-X100, 1.4 mM phenyl methyl sulfonyl fluoride (PMSF), 0.1% beta-mercaptoethanol (2-ME), 0.1 mg/mL lysosyme, 10 μg/mL DNaseI, and 10 µM ZnSO~4~). Cell disruption was induced by 5 rounds of sonication pulses (30% amplitude) for 30 seconds with a pause interval of 30 seconds. The samples were cooled to 4°C prior to sonication and kept on ice during sonication. The bacterial lysate was centrifuged at 10 000 rpm for 20 minutes at 4°C. PEI (polyethyleneimine, 0.8% v/v) was used to precipitate nucleic acid. The supernatant was subjected to affinity chromatography by incubation with glutathione-Sepharose 4B beads for 1.5 hours at 4°C. Subsequently, the beads were washed with 5 column volumes of wash buffer (150 mM NaCl, 50 mM Tris pH 7.2, and 1 mM DTT). Then, the GST-DOF-ZF was eluted using 20 mM Tris-HCl pH 7.2, 10 mM glutathione, 150 mM NaCl, 1 mM DTT. Fractions containing GST-DOF-ZF were collected and used in gel retardation assays. To prepare recombinant protein free from GST, the eluted GST-DOF-ZF protein was incubated with PreScission protease in cleavage buffer (150 mM NaCl, 50 mM Tris pH 7.2, 1 mM DTT) at 4°C overnight. The cleaved DOF-ZFs (which have 5 non-native amino acids at their N-terminus, of which 3 are derived from the cleavage site and 2 from the translation of the *Bam*HI site) was purified using UNO SI cation exchange chromatography (IEC) (BioRad). Proteins at each step of the expression and purification were subjected to SDS-PAGE analysis. Protein concentration was measured on a Nanodrop^®^ ND-1000 UV-Vis spectrophotometer (ThermoFisher Scientific, Wilmington DE, USA) at 280 nm, using extinction coefficients calculated from the amino acid sequence.
Atomic absorption spectroscopy {#s2-5}
------------------------------
The zinc content of the GST-DOF-ZF samples was determined by flame atomic absorption spectroscopy on a CTA3000 spectrometer (ChemTech Analytical Instruments Limited, UK). A protein sample was applied into the atomic absorption spectrophotometer and the signal monitored at the absorption maximum for Zn^2+^ (213.5 nm). Concentrations were determined in reference to a standard curve constructed using standard Zn^2+^ solutions (1000 mg/L in 0.5 mol/L nitric acid; BDH, stock solution diluted to final concentrations of 1-10 ppm). Two determinations were performed for each GST-DOF-ZF and the error value for the determination was calculated based on the error of the calibration curve.
Folding assessment of the proteins {#s2-6}
----------------------------------
Protein samples (100 µM) were prepared in 10 mM potassium phosphate buffer pH 7.2, containing 100 mM NaCl, 1 mM DTT, in 5% (v/v) D~2~O and 2 μM 2, 2-dimethyl-2-silapentane-5-sulfonate sodium salt (DSS). One-dimensional (1D) ^1^H NMR spectra were collected on a Bruker AvanceIII 600-MHz spectrometer and processed by TOPSPIN3 (Bruker, Karlsruhe, Germany). To record the fluorescence emission spectra of DOF-ZFs, protein samples were prepared at 2 μM concentration in a buffer comprising of 10 mM potassium phosphate buffer pH 7.2, 100 mM NaCl, and 0.1% 2-ME. The fluorescence intensity was recorded with a Jasco FP-750 spectrofluorimeter (Jasco Corporation, Tokyo, Japan) at 25°C, before and following the addition of guanidinium hydrochloride to a final concentration of up to 6 M. An excitation wavelength of 280 nm was used and the excitation slit width was set to 5 nm.
Oligonucleotide preparation {#s2-7}
---------------------------
The lyophilized 5'-fluorescein-labeled and unlabeled oligonucleotides were reconstituted to a concentration of 1 mM in DEPC water. For the preparation of dsDNA, forward and reverse oligonucleotides were mixed in a 1:1 molar ratio. The mixture was heated at 90°C for 10 min and then was cooled slowly to room temperature. The resulting double-stranded DNA was purified using a Superose™ 12 HR 10/30 column using an AKTA FPLC system (GE Healthcare Life Sciences, Silverwater, NSW). The concentration of each oligomer was determined by UV absorbance using their calculated extinction coefficient.
Gel retardation assay {#s2-8}
---------------------
Single- and double-motif probes ([Table 2](#T2){ref-type="table"}) were derived from the promoter sequence of the *GST6* and *DOF2.3* genes in *A. thaliana*. GST-DOF-ZF proteins (10 µM) were incubated with each dsDNA fragment (at 3 µM) in a total volume of 10 µL in a reaction buffer comprising 10 mM HEPES pH 7.8, 50 mM KCl, 5 mM MgCl~2~, 1 mM EDTA, 1 mM DTT and 5% glycerol. After a 30-minute incubation at 4°C, the samples were analyzed by 6% native polyacrylamide gel electrophoresis and DNA bands were visualized using ethidium bromide staining.
###### Oligonucleotide sequences used for gel retardation assays
-------------------------------- ------------------------------------------------ ----------------- ---------------
**Probe name** **Sequence** **Gene** **Reference**
Single-motif probe AATCC[AAAAG]{.ul}TGTAGAGGAG GST6 promoter Chen,1996
Double-motif probe AAATAATCAT[AAAG]{.ul}TATT[AAAG]{.ul}TAATATATAC DOF2.3 promoter Skirycz, 2008
Control for single-motif probe AATCCA[GCGA]{.ul}TGTAGAGGAG
Control for double-motif probe AAATAATCAT[GCGA]{.ul}TATT[GCGA]{.ul}AATATATAAC
-------------------------------- ------------------------------------------------ ----------------- ---------------
Microscale thermophoresis {#s2-9}
-------------------------
The affinity of DOF-ZF domains for target oligonucleotides was assessed using microscale thermophoresis. A constant quantity of fluorescently-labeled DNA (100 nM) was incubated with a range of concentrations of protein at room temperature for 15 minutes in 20 μL reaction buffer (10 mM potassium phosphate pH 7.2, 100 mM NaCl, 1 mM DTT, 5 mM MgCl~2~, 0.05% Tween 20). Samples were aspirated into the standard treated capillaries and scanned using a blue excitation laser on a Monolith™ NT.115 instrument (NanoTemper Technologies GmbH, München, Germany). The assessment was carried out using 50% LED power and 20% IR-laser power. The normalized fluorescence values were fitted to a simple 1:1 Langmuir binding isotherm in the Nanotemper software. The data from 3 individual measurements for each sample were used for the analyses.
Protein molecular weight determination using SEC and multiangle laser light scattering (SEC-MALLS) {#s2-10}
--------------------------------------------------------------------------------------------------
Purified proteins were subjected to size-exclusion chromatography using a Superose™ 12 HR 10/30 size-exclusion column (GE Healthcare, Parramatta, NSW) with an in-line MiniDawn MALLS detector with a laser source at 690 nm (Wyatt Technology, Santa Barbara, CA, USA) and Wyatt Refractometer. Proteins were eluted in 20 mM Tris pH 8.0, 150 mM NaCl, 1mM DTT using a flow rate of 0.4 mL/min. The weight-average molecular weight was calculated using the intensity of scattered light at 90° in combination with the change in refractive index. Protein concentration at the detector was determined by the change in refractive index.
Results {#s3}
=======
DOF domains are bona fide ZFs that form stable, folded structures {#s3-1}
-----------------------------------------------------------------
To investigate the interaction of DOF-ZF domains with DNA, the DOF domains from 3 DOF proteins from the model organism*A. thaliana*, namely DOF2.1, DOF3.4 and DOF5.8, were cloned into the pGEX-6p-1 bacterial expression vector. The constructs used here were 60 amino acids in length to cover the coding sequence of the predicted DOF-ZF domain. The similarity of these 3 DOF-ZF domains to other members of the family from different plant species is high, as is shown in [supplemental Fig. S1](#Suppl1){ref-type="supplementary-material"}. All 3 DOF-ZF domains were expressed at high levels in *E. coli* BL21 (DE3) as a fusion protein with a GST tag at the amino terminus and the GST-tagged DOF-ZF proteins were purified using glutathione affinity chromatography ([Fig. 2A](#F2){ref-type="fig"}). In order to assess whether these polypeptides are in fact zinc-binding domains, each of the 3 purified proteins was subjected to atomic absorption spectroscopy. The 3 proteins yielded Zn: protein ratios of 0.74±0.16 (DOF2.1-ZF), 1.07±0.08 (DOF3.4-ZF) and 1.07±0.23 (DOF5.8-ZF). Thus, these data are consistent with each DOF domain binding a single Zn^2+^ ion, most likely via the sidechains of the 4 cysteine residues highlighted in grey in [Fig. 1](#F1){ref-type="fig"}.
{#F2}
To allow further studies of these domains, the GST tag was removed by enzymatic cleavage and the DOF-ZF domains of DOF3.4 and DOF5.8 were successfully purified by cation exchange chromatography ([Fig. 2A](#F2){ref-type="fig"}). In contrast, DOF2.1-ZF underwent significant precipitation following cleavage and could not be obtained in appreciable amounts. As shown in [Fig. 2](#F2){ref-type="fig"}B, DOF3.4-ZF and DOF5.8-ZF were able to be prepared to a high level of purity (95%, as assessed by the image processing program ImageJ).
To assess the structural properties of these purified recombinant DOF-ZFs, we first recorded one-dimensional ^1^H NMR spectra for the DOF3.4-ZF and DOF5.8-ZF. The sharp and well-dispersed ^1^H NMR signals illustrated in [Fig. 3A](#F3){ref-type="fig"} suggest stably folded structures for both proteins. We also recorded fluorescence emission spectra of DOF3.4-ZF and DOF5.8-ZF to monitor the physical environment of the single tryptophan residue. [Fig. 3B and 3C](#F3){ref-type="fig"} show the variation in the intrinsic fluorescence emission spectra of both DOF3.4-ZF and DOF5.8-ZF, respectively, upon the addition of guanidinium hydrochloride (GndHCl) at concentrations ranging from 1 to 6 M. A significant redshift of λ~max~ (from 331 to 356 nm) was observed following the addition of the GndHCl, consistent with an increase in solvent exposure for the tryptophan. In 5 and 6 M GndHCl the spectra were identical with that of pure tryptophan (not shown), indicating complete denaturation of the protein. Therefore, it can be concluded that each domain takes up a well-ordered structure in phosphate buffer.
{#F3}
GST-DOF-ZFs bind AAAG motifs in a sequence-specific manner {#s3-2}
----------------------------------------------------------
We next investigated the ability of GST fused DOF-ZF domains to bind DNA. The promoter of the *GST6* gene, a gene involved in environmental stress responses, contains an octopine synthase (ocs) element and a DOF binding site upstream to the ocs element. In vitro gel retardation assays have shown that full-length DOF3.4 stimulates recruitment of OBF4 and OBF5 (ocs binding proteins) to their binding site and also that DOF3.4 binds by itself to a single AAAG sequence derived from the *GST6* promoter.^[@R15]^ We had synthesized a 20-base pair oligonucleotide that contains this single AAAG site from the *GST6* promoter (single-motif probe, [Table 2](#T2){ref-type="table"}). Another study using chromatin immunoprecipitation coupled to quantitative PCR verified the *DOF 2.3* gene promoter as another target of DOF3.4.^[@R16]^ The ability of the full -length DOF3.4 to the DNA fragment derived from the *DOF2.3* promoter was demonstrated using an in vitro gel retardation assay. This fragment contains 2 AAAG motifs in a parallel arrangement separated by 4 nucleotides and we selected this sequence for our experiments (double-motif probe, [Table 2](#T2){ref-type="table"}). We performed the gel retardation assays by titrating both single and double-motif probes with GST-DOF-ZF proteins. As shown in [Fig. 4](#F4){ref-type="fig"}, both probes exhibited interaction with all 3 GST-DOF-ZFs, albeit with significantly different characteristics leading to a decreased amount of free DNA at the bottom of the gel. All 3 double-motif probe-protein mixtures showed a clear retarded DNA band on the gel, whereas for single-motif probe there is no such discrete retarded band. This sharp band appears in a concentration-dependent manner, consistent with the formation of a single complex with a defined stoichiometry. The data suggest that the dissociation constant for formation of a complex with the double-motif probe under these conditions is \~3 µM for GST-DOF-2.1 and \~5 µM for GST-DOF-3.4 and GST-DOF-5.8.
{#F4}
To examine the possible contribution of GST dimerization to the observed bands, the DNA retardation assays were repeated for cleaved DOF3.4-ZF and DOF5.8-ZF using both single and double-motif probes ([Fig. 5](#F5){ref-type="fig"}). For the double-motif probe, binding is still clearly observed as a discrete band, although the shifted band is not as well defined as for the GST-fusion proteins. Inspection of the concentration dependence of binding indicates that the apparent dissociation constant is \~9 µM for both domains, which is within a factor of 2 of the estimated dissociation constants for the GST-fusion proteins. These data suggest that the presence of GST have a measurable but relatively small effect on binding of the DOF domains to DNA. The clearer shifted bands might indicate that the GST is somewhat decreasing the off-rate of the complex, increasing the lifetime of the complex as it runs through the gel.
{#F5}
In the case of the single AAAG probe, binding is also noticeably weaker although a smear is still visible, suggesting the formation of a complex with a faster off-rate. Moreover, to begin to assess the DNA-binding specificity of each domain, we also tested the binding of each DOF-ZF proteins with and without GST tag to oligonucleotide probes in which we had mutated AAAG motifs to GCGA ([Table 2](#T2){ref-type="table"}). None of the 3 DOF domains bound to the control oligonucleotides (data not shown), indicating that these domains are indeed targeting the AAAG motif.
To corroborate these binding data, we turned to microscale thermophoresis. Analysis of the interaction between 5'-fluorescein-labeled single-motif probe and either DOF5.8-ZF or DOF3.4-ZF revealed overall K~d~ values of greater than 100 µM, which is considered very weak for a DNA-binding ZF domain ([Fig. 6A](#F6){ref-type="fig"}). However, the DNA-binding affinity of DOF3.4-ZF and DOF5.8-ZF to the double-motif probe under the same conditions showed significantly tighter binding with overall K~d~ values of 2.3±1 µM and 2.5±1 µM, respectively ([Fig. 6B](#F6){ref-type="fig"}). Encouragingly, these values closely resemble the K~d~ estimates obtained from our gel retardation assays. To assess the oligomerization states of DOF-ZFs in solution, we used size exclusion chromatography coupled to multi-angle laser light scattering (SEC-MALLS). DOF3.4-ZF eluted from SEC-MALLS as a single species with a molecular weight of 6.9±4 kDa (theoretical Mw for the monomer = 7.3 kDa) ([Fig. 7A](#F7){ref-type="fig"}). As shown in Fig. 7B, the retention time of DOF5.8 is very similar to that of DOF3.4, indicating that this domain has the same self-association state as DOF3.4. These results show that both DOF domains are most likely in a monomeric state in solution in the absence of DNA.
{#F6}
{#F7}
Discussion {#s4}
==========
Previous in vivo and in vitro studies on DOF proteins have shown the ability of DOF-domain proteins to bind to specific promoters. Baumann et al showed that NtBBF1, a DOF protein from *Nicotiana tabacum*, binds to ACTTTA site in the *rolB* promoter and activates the expression of the beta-glucuronidase reporter gene in the apical meristem and vascular system of transgenic tobacco.^[@R17]^ In another study, full-length StDOF1 from potato was shown in gel retardation assays to bind to a DNA fragment containing 3 TAAAG repeats that is found in the promoter of the *KST1* gene (which encodes a K^+^ influx channel that is expressed predominantly in guard cells.^[@R18]^ However, until now there have been no reports in the literature of the quantitation of direct DNA binding of isolated DOF-ZF domains.
In this work, N-terminally GST-tagged polypeptides encoding the putative DOF-ZF domains from *A. thaliana* DOF2.1, DOF3.4, and DOF5.8 were expressed in *E. coli* and purified either as fusion proteins or as isolated DOF domains. The final proteins were purified to \~95% judged based on SDS-PAGE analysis. Examination of the DOF-ZF domains by ^1^H NMR spectroscopy revealed that the DOF-ZF domains were able to form a well-defined 3-dimensional structure and atomic absorption spectrometry confirmed that all of the DOF domains bind the expected one molar equivalent of Zn^2+^.
We went on to carry out gel retardation assays using (a) a single-motif DNA probe, from the *GST6* promoter, which contains a single AAAG motif, and (b) a double-motif probe derived from the *DOF2.3* promoter, which contains 2 AAAG motifs. All 3 GST-DOF-ZF proteins interacted with both oligonucleotides. However, only the complexes formed with the double-motif probe were observed as well-defined bands in the gels; the single-motif probe complexes were highly smeared, suggesting that the double AAAG motif forms higher affinity complexes.
To assess these interactions in more detail, we used MST to measure the dissociation constants for the binding of the double-motif probe to the (GST-free) ZF domains of DOF3.4 and DOF5.8; in both cases, a K~d~ of \~2 µM was obtained. In contrast, single-motif probe displayed very weak binding to both DOF domains (\>100 µM). Collectively, the results of gel retardation assay and MST experiments suggest the presence of only a very weak interaction between the single-motif probe and DOF-ZFs furthermore, the DOF-ZF domain affinity for DNA is significantly stronger (at least 100-fold) when the target DNA contains 2 repeats of the AAAG motif. This conclusion is in agreement with gel retardation data for full-length His~6~-DOF3.4 binding to the promoter of *DOF2.3* from *A. thaliana*.^[@R16]^
The stronger interaction observed for the double-motif probe could arise from one of 2 possible scenarios. First, one DOF-ZF could bind to each site and the 2 ZFs could contact each other, stabilizing their interaction with DNA. Although no appreciable self-association is observed for DOF3.4 or DF5.8 ZFs in solution, it is quite plausible that they could make contacts when localized by their binding to the same DNA probe. The second possibility is that the context of the AAAG motifs in the double-site probe is sufficiently different from the single-motif probe that they account for the \~100-fold increase in binding affinity. Overall, we consider the first scenario to be more likely (and more consistent with published data on the homo- and heterodimerization of DOF proteins; see below), and additional confirmation awaits detailed structural studies of these proteins bound to DNA.
There are numerous reports describing nanomolar (or stronger) affinities of ZF proteins for DNA.^[@R19],[@R20]^ Generally, an array of 3 or more ZFs is required for tight, specific binding of this type, although all ZFs in a protein may not be equally involved in protein or DNA interactions.^[@R21]^ However, there are examples of ZF proteins that tightly interact with their cognate DNA through a single ZF domain. For example, the DNA affinity of the one-ZF GAGA protein from *Drosophila melanogaster* is in the nanomolar range. However, in this case, 2 basic regions N-terminal to the GAGA ZF domain are thought to make interactions with the DNA.^[@R22]^ Several other one-ZF DNA-binding proteins make use of an adjacent region that has some affinity for nucleic acids. For example, the TFs NIT2 from *Neurospora crassa* and AreA from *Aspergillus nidulans* each bear a single Cys4 ZF and display a C-terminally extended basic region that contributes to DNA binding ([Fig. 8](#F8){ref-type="fig"}).^[@R23],[@R24]^ The affinity of DOF-ZF domains for the double-motif probe measured in this study is comparable to the affinity of the minimal ZF domain of AreA (lacking the basic region) for its cognate DNA site, in contrast to the sub-nanomolar affinity measured for a C-terminally extended version of the domain. These extended regions can potentially contribute to DNA-binding affinity through direct interactions (either specific or nonspecific) with DNA or by stabilizing the structures -- such as ZF domains -- that are directly involved in the binding.^[@R24]^ The constructs used in the current study did not include the extended C-terminal basic region in the DOF-ZF sequences.
{#F8}
Genome-wide analysis of *A. thaliana* reveals a high frequency of repeated DOF binding sequences in the same or inverted orientations, with the most common arrangement being AAAG-N~7~-CTTT. In this study, it was shown that the predominant bases separating AAAG sequences in AAAG-Nn-AAAG pattern are A and G (N denotes any nucleotide bases).^[@R25]^ Such repeats are found in the promoters of the *ANAC069* (seven copies in both orientation) and *MYB60* (a cluster of DOF motifs in both orientation) genes, which encode a member of NAC family of plant-specific TFs (NTM2) and a guard-cell specific protein, respectively.^[@R26],[@R27]^ This feature is not limited to the *A. thaliana* as these repeated elements are abundantly seen in the promoters of genes controlled by DOF proteins in other plants. Guard cell specific KST1, ADP-glucose pyrophosphorylase, and PEP carboxylase genes in potato each contain multiple copies of such a motif.^[@R18]^ The position of these motifs in the promoter region and the number of repetitions and of intervening nucleotides differ between genes, but these repeated motifs might give rise to higher affinity and/or higher specificity DNA binding.
Interactions between TFs are important in the regulation of gene expression. The interaction of DOF proteins with each other and with different TFs has been observed previously. A previous gel retardation analysis of DOF1 and DOF2, 2 maize DOF proteins, demonstrated that these proteins are able to self-associate and create homomeric and heteromeric complexes.^[@R28]^ The non-tagged DOF-ZF domain of DOF1 was shown to specifically interact with DNA probes containing one and 2 binding sites (separated by 7 nucleotides). The DOF1-ZF formed higher order complexes with both DNA probes as its concentration increases. This could be due to the binding of dimerized DOF1-ZF to the binding site.^[@R28]^ Other studies have shown that particular DOF proteins need an 'assistant' protein for optimum binding to DNA. In barley, in vivo studies revealed the physical interaction between HvWRKY38 and BPBF (a DOF protein), as a repressor, in regulating the gibberellin induced Amy32b α-amylase protein. The binding sites of these 2 TFs are only 14 bp apart.^[@R29]^ Finally, in vitro experiments using gel retardation assays indicated that the maize HMGB protein stimulates the binding of the DOF-ZF domain of the maize TF, DOF2, to a probe with a single AAAG site. In the absence of the HMGB protein, the amount of complex formed between DOF2-ZF and DNA reduced significantly.^[@R30]^
Overall, it can be concluded that self-association and formation of the protein complexes are likely to be mechanisms that DOF proteins use for tight binding to their target sites. In the present study, the binding affinity of DOF3.4-ZF and DOF5.8-ZF to the probe with a single binding site was very weak compared to their affinity to the probe with 2 binding sites. Since the binding sites are only 4 bp apart, tighter binding for 2 sites possibly indicates that proteins are able to physically contact each other and bind as a dimer to the DNA.
Conclusion {#s5}
==========
In summary, this work presents experimental evidence that DOF-ZF domains are bona fide ZF domains and that the DOF-ZF domains in isolation bind to DNA fragments containing AAAG sequence. Quantitative assessment of the binding is also performed. According to our results, the binding affinity of DOF-ZF domains to an oligonucleotide containing 2 binding sites is \~100-fold higher than that for a similar oligonucleotide with one binding site. This observation provides a possible explanation for the presence of repeated numbers of AAAG in the promoter regions of DOF TF target genes.
Ethical approval {#s6}
================
None to be declared.
Competing interests {#s7}
===================
There is no conflict of interests to be reported.
Supplementary Materials {#s8}
=======================
######
Supplementary file 1 contains Fig. S1.
######
Click here for additional data file.
Acknowledgment {#s9}
==============
The authors would like to thank the Research Office of Tabriz University of Medical Sciences for providing financial support under the Postgraduate Research Grant scheme toward the Ph.D. thesis No: 92/4-5/2. The support of School of Life and Environmental Sciences, The University of Sydney is also gratefully acknowledged.
Research Highlights {#s10}
===================
What is current knowledge? {#research1}
--------------------------
1. √ DOF proteins are plant-specific TFs.
2. √ Zinc finger domains of DOF proteins from *A. thaliana* bind to their specific AAAG cognate DNA with unknown affinity.
What is new here? {#research2}
-----------------
1. √ DOF-ZFs are structurally folded in the monomeric state in the solution.
2. √ The binding affinity of DOF-ZF domains to their cognate DNA sequence was determined using biophysical methods.
3. √ A mechanism for binding of DOF-ZFs to DNA was proposed.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Overnutrition is defined as excess nutrient supply for normal metabolism and growth. Overnutrition has been linked to both insulin (INS) resistance and *β*-cell dysfunction \[[@B1], [@B2]\]. The global obesity epidemic is a major reason for the increased incidence of type 2 diabetes mellitus (T2DM) and associated cardiovascular diseases (CVD). Obese individuals may have CVD complications, including hypertension, cardiac diastolic dysfunction, macro- and microvascular disease, and early onset of cardiovascular morbidity \[[@B3], [@B4]\].
Chronic overnutrition leads to activation of the evolutionarily conserved mammalian target of rapamycin (mTOR) and its downstream signaling molecule, p70 ribosomal S6 kinase 1 (S6K1) \[[@B5]\]. Activation of the INS receptor triggers phosphorylation of tyrosine residues and subsequent activation of the INS docking proteins (INS receptor substrates 1 and 2; IRS-1, IRS-2). It results in the phosphorylation/activation of phosphatidylinositol 3-kinase and protein kinase B (Akt). Tyrosine phosphorylation of IRS-2 appears to play an important role in the antiapoptotic actions of INS in *β*-cells \[[@B6]\]. mTOR/S6K1 signaling impairs INS metabolic signaling through enhanced serine (Ser) phosphorylation of IRS-1, a mechanism implicated in the pathogenesis of INS resistance. Ser phosphorylation of IRS-1 and IRS-2 result in their targeting for proteosomal degradation, which, in turn, leads to impaired INS metabolic signaling \[[@B7]\]. Dipeptidyl peptidase-4 (DPP-4) inactivates and degrades glucagon-like peptide- (GLP-) 1 and glucose-dependent insulinotropic polypeptide (GIP) \[[@B8]\], thereby limiting the favorable metabolic action of these proteins. DPP-4 inhibitors such as sitagliptin (SIT) are a new therapeutic strategy for improving glucose metabolism in patients with T2DM.
SIT inhibits DPP-4, which is responsible for degrading GLP-1. Inhibition of DPP-4 increases circulating levels of GLP-1. GLP-1 is produced by L-cells in small and large intestines in response to nutrient intake \[[@B9]\]. GLP-1 is degraded by DPP-4 in the bloodstream \[[@B9]\]. SIT also regulates GLP-1 receptor and cannabinoid receptor- (CB-) 1 gene expressions, which are associated with appetite regulation in diabetic rat and may decrease oxidative stress and liver tissue damage \[[@B10]\].
Female mice showed that they are more prone to develop obesity than male mice. An 8-week administration of Western diet (WD) abolished the enhancement of insulin sensitivity and induced cardiac diastolic dysfunction in female mice but not in male mice \[[@B11]\]. SIT improves glucose metabolism and insulin sensitivity in both human and animal models \[[@B12], [@B13]\]. SIT also attenuated the annual exacerbation of diastolic dysfunction in patients with T2DM for 24 months \[[@B14]\]. However, the mechanism underlying this protective effect of SIT remains unclear. In this study, we used the overnutrition obesity female mouse model to evaluate whether a 12-week treatment with SIT improves cardiac diastolic dysfunction and glucose homeostasis and reduces *β*-cell dysfunction. We hypothesized that the SIT may exert its effect through insulin metabolic signaling molecule, p70 ribosomal S6K1, and through phosphorylation of INS docking proteins IRS-1 and IRS-2.
2. Methods {#sec2}
==========
2.1. Animals {#sec2.1}
------------
The experimental study was approved by the Suzhou Science and Technology Town Hospital Institutional Animal Care and Use Committee. Six-week-old wild-type control (C57BLKS/J) female mice were purchased from the animal center of Soochow University (Suzhou, China) and were housed under standard laboratory conditions. Groups of 6-week-old mice were fed a WD consisting of high-fat (46%) and a high-carbohydrate component as constituted with sucrose (17.5%) and high-fructose corn syrup (17.5%) and water for 12 weeks with or without SIT (15 mg/kg/day) \[[@B15]\]. Another group of age-matched controls were fed regular mouse chow (CD) with or without SIT for the same period of time.
2.2. Biochemical Examination {#sec2.2}
----------------------------
Venous blood sample was collected from fasting mice for four hours. Glucose and insulin assay were analyzed using an alpha track II glucometer and an ultrasensitive mouse insulin ELISA kit (Crystal Chemical Inc., Wakefield, MA, United States). Samples were drawn immediately prior to the start of the treatment period and at the end of the study. Hemoglobin A1c (HbA1c) was measured using a DCA vantage analyzer (Seimens City, United States). Plasma total cholesterol and triglycerides, serum potassium, and sodium were measured using an Olympus AU680 automated chemistry analyzer (Beckman Coulter, Brea, CA). Triglyceride content in liver samples was also analyzed \[[@B16]\].
2.3. Echocardiography {#sec2.3}
---------------------
Two-dimensional echocardiograms were performed as described previously \[[@B11]\]. The myocardial performance index (MPI) was calculated as the sum of isovolumic contraction and relaxation times divided by ejection time. The data was obtained by a pulsed-wave Doppler, and parameters were assessed. Calculations were made in accordance with the specific guidelines for rodent echocardiography. All data were acquired and analyzed by a single-blinded observer using Echo PAC (GE Vingmed) offline processing.
2.4. INS Signaling Protein Assay {#sec2.4}
--------------------------------
Pancreas islet cells will be isolated by type V collagenase digestion, followed by Ficoll 400 gradient separation, as described in \[[@B17]\]. Proteins involved in INS signaling, including IRS-1, IRS-2, Akt, and S6K1, will be analyzed in pancreatic islets by Western blot.
2.5. Hematoxylin and Eosin Staining {#sec2.5}
-----------------------------------
Specimens of the hearts were fixed by formalin for 24 hours, dehydrated by 70%, 80%, and 90% ethanol for 3 hours, respectively, then 100% ethanol I for 2 hours and 100% ethanol II for 2 hours, and vitrified by xylene I and xylene II for 20 minutes. After immersing in paraffin I and II for 40 minutes, the specimens were embedded and sliced (5 *μ*m). Staining was performed as follows: hematoxylin staining for 15 minutes, hydrochloric acid alcohol solution for 35 seconds decoloring, eosin staining for 10 minutes, and 90% ethanol for 40 seconds decoloring. Then, neutral balsam was used for mounting, and the section was observed and photographed under the microscope \[[@B18]\].
The heart injuries were divided into three categories. Normal structure: the myocardial structure is normal, the cells are closely aligned, the boundaries are clear, and there is no obvious degeneration, congestion, and edema. Moderate structure injury: the myocardial arrangement was slightly irregular, the myocardial cells were swollen, the myocardial hyperemic edema was visible, and a small amount of lymphocyte infiltration was visible. Serious structure injury: The injury shows loose arrangement of myocardial cell, irregular part of the myocardial fiber fracture, apparent vacuole degeneration, height of myocardial interstitial hyperemia and edema, and a large amount of inflammatory cell infiltration.
2.6. Statistical Analysis {#sec2.6}
-------------------------
All data are expressed as means ± SE. A one-way analysis of variance was used to assess the overall difference between groups. A value of *P* \< 0.05 was considered statistically significant.
3. Results {#sec3}
==========
3.1. Baseline Parameters {#sec3.1}
------------------------
In this study, body weight over the course of the 12-week treatment period was elevated in both CD and WD groups. In addition, body weight in the WD group increased more significantly than that in the CD group (*P* \< 0.01, [Table 1](#tab1){ref-type="table"}). SIT treatment significantly reduced the elevation of body weight in the WD + SIT group (*P* \< 0.05). Body weight and % of body weight gain did not differ between CD and CD + SIT groups after treatment (*P* \> 0.05).
3.2. Metabolic Parameters {#sec3.2}
-------------------------
Liver weight, plasma cholesterol, triglycerides, and alanine aminotransferase (ALT) were similarly elevated in the WD group, indicating lipidemia and liver impairment in WD mice ([Table 1](#tab1){ref-type="table"}). Liver weight elevation in WD mice was not reduced by SIT treatment. SIT treatment decreased the elevation of cholesterol, triglycerides, and ALT in WD mice (*P* \< 0.05).
3.3. Glucose Homeostasis Parameters {#sec3.3}
-----------------------------------
Baseline fasting glucose and HbA1c in all groups did not show significant difference at the beginning of the study. At the end of study, WD groups had higher elevated fasting glucose and HbA1c compared to CD groups (Figures [1(a)](#fig1){ref-type="fig"} and [1(b)](#fig1){ref-type="fig"}), and the fasting glucose and HbA1c in the WD + SIT group were significantly lower compared to those in the WD group (*P* \< 0.05). However, the values did not return to the normal range. The results indicate that progressing hyperglycemia was ameliorated by SIT. In this study, mice in the WD group had higher plasma insulin concentrations compared to CD group mice (*P* \< 0.01). SIT also reduced the plasma insulin significantly in the WD + SIT group compared to that in the WD group ([Figure 1(c)](#fig1){ref-type="fig"}).
3.4. WD-Induced Diastolic Dysfunction Was Improved by SIT {#sec3.4}
---------------------------------------------------------
Female mice fed with WD for 12 weeks exhibited abnormal echocardiographic diastolic function parameters when compared with CD and CD + SIT groups (Figures [2(a)](#fig2){ref-type="fig"}--[2(d)](#fig2){ref-type="fig"}). The MPI, which assesses both systolic and diastolic function, was increased in the WD group, but its impact was greatly decreased by SIT treatment ([Figure 2(a)](#fig2){ref-type="fig"}). An increase in this parameter is indicative of impaired cardiac function. The increase in MPI in the WD group is likely due to abnormal diastolic function as indicated by both a prolonged period of isovolumic relaxation ([Figure 2(b)](#fig2){ref-type="fig"}) and a decrease in mitral inflow Vp ([Figure 2(c)](#fig2){ref-type="fig"}). The *E*/Vp ratio, a marker of left ventricle filling pressure, was elevated in the WD group compared with CD and CD + SIT groups. This further supports the diagnosis of diastolic dysfunction ([Figure 2(d)](#fig2){ref-type="fig"}). SIT treatment effectively decreased this elevation in filling pressure. SIT treatment also significantly ameliorated the impact effects of WD-induced cardiac diastolic function.
3.5. WD-Induced Abnormalities in Myocardial Structure Were Improved by SIT {#sec3.5}
--------------------------------------------------------------------------
The myocardial structure in the CD group and CD + SIT group exhibited normal histology. The cells are closely aligned with clear boundaries. They also do not show degeneration, congestion, and edema (Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}). Compared to CD mice, WD mice showed severe myocardial structure injury ([Figure 3(c)](#fig3){ref-type="fig"}). In the WD + SIT group, hearts showed a moderate myocardial structure injury. Overall, there is less evidence of cellular injury than in the WD group ([Figure 3(d)](#fig3){ref-type="fig"}). The injury of myocardium structure was less severe in WD + SIT group mice than that in WD group mice.
3.6. Evaluation of Insulin Signaling in Pancreas {#sec3.6}
------------------------------------------------
In WD fed mice, progressive *β*-cell failure leads to overt hyperglycemia. They usually develop obesity and hyperglycemia due to endocrine pancreatic insufficiency \[[@B19]\]. Western blot evaluation demonstrated decreased total IRS-1 and IRS-2 and an increased phosphorylated Ser^636^ of IRS-1 and Ser^731^ of IRS-2 in pancreas tissue (Figures [4](#fig4){ref-type="fig"} and [5](#fig5){ref-type="fig"}). SIT treatment reduced the degradation of IRS-1 and IRS-2 (*P* \< 0.05) and phosphorylation of IRS-1 and IRS-2. S6K1 phosphorylation was increased in WD fed mice compared to CD counterparts and was decreased by SIT treatment ([Figure 6](#fig6){ref-type="fig"}). Our pancreas tissue data confirmed *β*-cell dysfunction in the WD fed mouse and demonstrated excessive activation of S6K1 as well as increased abnormal Ser phosphorylation of IRS-1 and IRS-2, along with degradation of total IRS-1 and IRS-2. Moreover, these alterations were attenuated by SIT treatment.
4. Discussion {#sec4}
=============
The aim of the study was to determine whether a 12-week treatment with DPP-4 inhibitor SIT ameliorates overnutrition-induced progression of abnormal cardiac dysfunction and glucose homeostasis in obese female mice. The role of insulin metabolic signaling molecules was also investigated. This study showed that DPP-4 inhibitor SIT improves cardiac diastolic function in female obese mice. This improvement was associated with the protective effect of *β*-cell function by reductions of mTOR/S6K1 activation, degradation, and serphorylation of INS docking proteins IRS-1 and IRS-2. The results show that targeted pharmacologic interventions with DPP-4 inhibitor could be useful in ameliorating pathophysiologic abnormalities in cardiac diastolic dysfunction and preventing the activation of insulin metabolic signaling molecules through enhanced mTOR/S6K1 activation and Ser phosphorylation of IRS-1 and IRS-2. This is one mechanism implicated in the pathogenesis of INS resistance in the setting of obesity or diabetes.
Chronic overnutrition with a WD resulted in obesity, insulin resistance, and elevated plasma DPP-4 activity as well as heart enlargement and dysfunction \[[@B20]\]. Diabetic db/db mice are reported to exhibit increased interstitial fibrosis as early as 2 months of age \[[@B21]\]. We observed an improvement in cardiac diastolic function which was associated with reductions in myocardial fiber fracture, apparent vacuole degeneration, myocardial interstitial hyperemia, edema, and inflammatory cell infiltration in WD fed mice with SIT administration. Cardiac structure in the WD group was more impaired, and this could be due to the result of stress on the left ventricle wall caused by an increase in left ventricle filling pressure. DPP-4 inhibitor linagliptin suppresses reactive oxygen species (ROS) generation, nicotinamide adenine dinucleotide phosphate (NADPH) oxidase, and proinflammatory signals and reduces collagen deposition \[[@B22]\]. There is growing evidence that a DPP-4 inhibitor could exert cardioprotection and improve left ventricular function by reducing oxidative stress and apoptosis and increasing reperfusion injury salvage kinase (RISK) activity \[[@B23]\]. Thus, it is likely that SIT may blunt myocardium injury progression. The underlying mechanism needs to be further investigated.
Multiple metabolic and proliferative pathways implicated in INS resistance lead to Ser phosphorylation of IRS-1. However, the role of the mTOR/S6K1 signaling pathway is particularly interesting because it is affected by nutrients and energy status. Studying this pathway allows for a comprehensive evaluation of the role of overnutrition on INS resistance and *β*-cell failure. Excessive activation of mTOR/S6K1 can also impair INS metabolic signaling through enhanced Ser phosphorylation of IRS-1 and IRS-2, which targets these molecules for proteosomal degradation. This is a widely accepted mechanism which contributes to the pathogenesis of INS resistance in several tissues, including skeletal muscle \[[@B24]--[@B26]\]. Our data in the pancreas tissue confirmed *β*-cell dysfunction in the WD fed mouse and demonstrated excessive activation of S6K1 as well as increased abnormal Ser phosphorylation of IRS-1 and IRS-2, along with degradation of IRS-1 and IRS-2. Moreover, these alterations were attenuated by SIT treatment in our obese female mouse model. Other studies have demonstrated increased mTOR/S6K1 activation in the transgenic Zucker obese (ZO) rat \[[@B2], [@B27]\], a model of overnutrition and obesity which carries a mutation of the leptin receptor and develops INS resistance and glucose intolerance. This suggested that the mTOR/S6K1 and INS metabolic signaling pathways in the pancreas play an important role in the development and survival of *β*-cells. S6K1 KO mice have decreased *β*-cell size and mass, hypoinsulinemia, and glucose intolerance, suggesting a critical participation of this pathway in *β*-cell survival \[[@B28]\]. How DPP-4 inhibitor attenuates the activation of the S6K1 in the setting of overnutrition- and obesity-induced INS resistance as well as its relation to *β*-cell function is yet fully elucidated.
Chronic treatment of WD fed mice with the DPP-4 inhibitor led to marked inhibition of plasma DPP-4 activity and improved insulin sensitivity. DPP-4 inhibition may suppress INS resistance with a decrease in body weight but not back to control levels \[[@B20]\]. In a fatty liver Shionogi-ob/ob male mouse model, SIT administration reduced body weight, blood glucose levels, and hepatic fibrosis. It also attenuated hepatic stellate cell activation and Kupffer cells \[[@B29]\]. Consistent with these studies, we observed the differences in body weight between CD and WD at the end of experiment in our female obesity model. However, in insulin-resistant male ZO rats, DPP-4 inhibitor linagliptin treatment for 8 weeks did not alter the body weight during the study period \[[@B30]\]. This difference in results may due to the different animal strain used (rat versus mouse; male versus female), as well as different inhibitor and treatment time.
In our study, food intake or calorie intake was not measured to evaluate whether beneficial effects of SIT could be due to a decrease in food intake or low-calorie intake. A research study demonstrated that SIT attenuated body adiposity, without affecting food intake, in C57BL/6 mice with diet-induced obesity \[[@B31]\]. However, in a clinical study patients who were given with SIT 100 mg (oral) daily for 4 weeks, postprandial serum glucagon, fasting blood glucose, and 24-h caloric intake decreased \[[@B32]\]. The possibility that SIT exerts its beneficial effects in part via suppression of food intake or low-calorie intake cannot be excluded.
Saxagliptin monotherapy prevented or delayed the progression of impaired glucose tolerance or impaired fasting glucose to type 2 diabetes mellitus in obese patients with newly diagnosed prediabetes \[[@B33]\]. SIT administration decreased ambient blood glucose levels and improved the glucose excursion rate. This was associated with elevated plasma insulin and reduced plasma glucagon levels \[[@B34]\]. SIT also decreased circulating DPP-4 activity, improved glucose tolerance, glucose-stimulated insulin secretion, and insulin sensitivity, and reduced plasma triglycerides and cholesterol levels \[[@B35]\]. In the recent study, SIT protected liver tissue, modulated lipid metabolism in a mouse model, and mediated expression levels of key enzymes for lipid metabolism \[[@B15]\]. Our results also showed that SIT decreased the elevation of cholesterol, triglycerides, and ALT in the WD group. Given that SIT altered the elevation of cholesterol, triglycerides, and ALT in the WD + SIT group compared to the WD group, this indicates that plasma glucose metabolic change may be involved in the improvement effect of SIT on cardiac dysfunction in WD fed female mice. In addition, high-fat fed mice treated with SIT exhibited significant improvement in insulin sensitivity indicating that improvements in glycemic control were not solely a consequence of enhanced insulin secretion \[[@B36]\]. The recent study suggested that the protective effects afforded by this DPP-4 inhibitor may derive from improvement of the metabolic profile and from cytoprotective properties \[[@B37]\]. In the Zucker diabetic fatty (ZDF) rat, Ferreira et al. have found that chronic SIT treatment corrected the glycemic dysmetabolism, hypertriglyceridemia, inflammation, and hypertension and reduced the severity of the histopathological lesions of pancreatic endocrine and exocrine tissues, together with a favorable redox status, which provides a further advantage in the management of diabetes and its proatherogenic comorbidities \[[@B38]\].
In summary, our study supports a newly described pleiotropic protective effect of DPP-4 inhibitor SIT on diastolic function and *β*-cell function in the obesity female mouse model. Despite the improvement in glycemic control, the HbA1c and fasting glucose values remained elevated. It is likely that the pleiotropic effects of SIT relate to factors other than improvements in glycemia and lipidemia. These findings suggest a potential clinical utility for SIT in the obesity/diabetic population. However, some limitations exist in the current study, such as pancreas histomorphology could not be done to study the impact of SIT on pancreas lesions and the indirect effect on dysfunction by assessing S6K1, IRS-1, and IRS-2 was not evaluated. Additional studies are needed to further elucidate the potential role of mTOR/SGK1 and INS docking proteins INS receptor substrates 1 and 2 as mediators of the efficacy of SIT on insulin resistance. DPP-4 activity and GLP-1 levels will be also needed to be measured to confirm that chronic overnutrition elevates plasma DPP-4 activity and DPP-4 inhibitor inhibits plasma DPP-4 activity.
This work was supported by Grants SS201756 (to Dr. Jianzhong An) and SS201613 (to Dr. Chen Wang) from Suzhou Science and Technology Development Plan, NSFC 81703501 (to Dr. Shigang Qiao) from the National Science and Technology Development Plan, QNRC2016219 (to Dr. Shigang Qiao) from Jiangsu Key Talent Youth Awards in Medicine, Gusu Health Youth Talent Awards (to Dr. Shigang Qiao), and 2017Q003 (to Lei Hong) from Suzhou New District Science and Technology Project.
Data Availability
=================
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
=====================
The authors declare that they have no competing interests.
Authors\' Contributions
=======================
Shigang Qiao and Guofang Mao contributed equally to this work.
{#fig1}
{#fig2}
{#fig3}
{#fig4}
{#fig5}
{#fig6}
######
Baseline and posttreatment parameters of body weight and plasma metabolic markers in control and WD group mice with or without SIT treatment. Values are mean ± SE. ^∗^*P* \< 0.05 compared to the CD or CD + SIT group; ^§^*P* \< 0.05 compared to the WD group. Sample sizes are noted in parentheses.
Parameter CD (12) CD + SIT (12) WD (12) WD + SIT (12)
--------------------------------------- ------------ --------------- --------------- -----------------
Pretreatment body weight (g) 18.1 ± 0.3 18.2 ± 0.4 18.1 ± 0.3 18.1 ± 0.4
Posttreatment body weight (g) 28.3 ± 1.3 28.8 ± 1.5 51.6 ± 2.5^∗^ 44.5 ± 2.6^∗,§^
Hepatic weight (mg) 826 ± 32 819 ± 39 2453 ± 196^∗^ 2232 ± 188^∗^
Hepatic triglycerides (nmol/g) 15 ± 3 16 ± 3 45 ± 4^∗^ 32 ± 4^∗,§^
Plasma cholesterol (mg/dL) 82 ± 5 80 ± 6 157 ± 6^∗^ 132 ± 8^∗^
Plasma triglycerides (mg/dL) 133 ± 10 137 ± 11 325 ± 16^∗^ 269 ± 18^∗,§^
Plasma alanine aminotransferase (U/L) 25 ± 2 27 ± 3 82 ± 6^∗^ 63 ± 5^∗,§^
[^1]: Academic Editor: Rosa Fernandes
| {
"pile_set_name": "PubMed Central"
} |
Findings {#Sec1}
========
Background {#Sec2}
----------
*Toxoplasma gondii* is a ubiquitous protozoan parasite capable of infecting all warm-blooded animals, including humans \[[@CR1]\]. The definitive host, the cat, passes oocysts in its faeces that contaminate the environment where they can remain viable for long periods of time. Intermediate hosts, such as livestock, are infected by ingesting contaminated water or soil and this results in the formation of tissue cysts, containing *T. gondii,* that can remain viable for the lifetime of the host \[[@CR2]\]. Although most infections in small ruminants are asymptomatic there can be abortions, foetal mummification, stillbirths and birth of weak lambs/kids \[[@CR3]\]. Humans can become infected with *T. gondii* by ingesting tissue cysts from meat, or oocysts from contaminated food or water. Most infections are asymptomatic; however, there can be severe neurological and pulmonary signs in immunosuppressed people. Infections of a foetus during pregnancy can result in retinochoroiditis and neurological abnormalities \[[@CR4]\].
Small ruminants are a very important source of domestic meat production in the Caribbean \[[@CR5]\]. Generally, the Caribbean is a net importer of meat, particularly beef and poultry; however, the demand for sheep and goat meat (both referred to as mutton) is high, and any meat produced locally is consumed by the local population and occasionally by foreign tourists interested in tasting local cuisine \[[@CR5]\].
The prevalence of *T. gondii* in food animals is higher in pigs, sheep and goats than in cattle \[[@CR6], [@CR7]\], and varies worldwide \[[@CR8]\]. There are few studies on the prevalence of *T. gondii* in livestock animals in the Caribbean. In Trinidad, a study of animals at slaughter reported *T. gondii* antibodies in 42.9% of goats and 5.5% of pigs \[[@CR9]\]. In a more recent study, 23.1% of pigs, 44.1% of sheep, 42.8% of goats and 8.4% of cattle in Grenada and Carriacou were seropositive for *T. gondii* \[[@CR10]\].
To provide further data on the epidemiology of *T. gondii* in small ruminants in the Caribbean, we tested sheep and goats from four different islands for reactive antibodies.
Methods {#Sec3}
=======
Sera used in the study had previously been obtained from mixed-breed sheep and goats on the islands of Dominica (15° 25' North, 61° 20' West), Grenada (12° 07' North, 61° 40' West) Montserrat (16° 45' North, 62° 12' West) and St. Kitts and Nevis (17° 20' North, 62° 45' West) between 2007 and 2012 and stored at −80°C \[[@CR11], [@CR12]\]. All samples were convenience samples collected from healthy, accessible animals on each island, as part of two studies investigating the presence of *Rickettsia africae* \[[@CR11]\] and *Ehrlichia ruminantium* \[[@CR12]\] in domestic ruminants. The animals were free to graze on open pasture during the day and were housed at night in open-air pens. The climate across the Caribbean is typically tropical with little variation in the coastal areas where the sampling sites were located. Ethical approval for the collection of the blood samples was obtained from the Institutional Animal Care and Use Committee, Ross University School of Veterinary Medicine, St. Kitts. Sera were examined for *T. gondii* antibodies using an in-house ELISA \[[@CR13]\], with modifications. In brief, 96-well microtitre plates were coated overnight with 3 μg/ml solubilised RH antigen \[[@CR14]\], washed with PBST (PBS with 0.05% Tween-20) and incubated for 2 h at 37°C after addition of 100 μl test or control sera (diluted 1:500 in 1% BSA in PBST) per well. Following washing, 100 μl HRP-conjugated Protein G (diluted 1:20,000 in 1% BSA in PBST) was added to each well and plates incubated for 2 h at 37°C. ELISAs were developed with TMB and reactions stopped with 2 M H~2~SO~4~ before reading ODs at 450 nm. Control sera were pooled samples of 5 sheep experimentally infected with *T. gondii*, and 5 negative control sheep from the same experiment \[[@CR15]\]. For each plate, the cut-off value was calculated as two times the percent positivity of the negative control serum relative to the positive control serum (i.e. \[2 x (average negative control sera OD/average positive control sera OD)\] x average positive control) \[[@CR16]\].
Results {#Sec4}
=======
Antibodies to *T. gondii* were detected in sheep and goats, respectively, from Dominica (67%, 37/55; 58%, 79/136), Grenada (48%, 40/84; 57%, 54/94), Montserrat (89%, 25/28; 80%, 25/31) and St. Kitts and Nevis (57%, 78/138; 42%, 76/181) (Table [1](#Tab1){ref-type="table"}).Table 1**Seroprevalence of** ***Toxoplasma gondii*** **in small ruminants in the Caribbean**IslandSheep tested (n)Sheep positive (n)% Sheep positiveGoats tested (n)Goats positive (n)% Goats positiveDominica553767.31367958.1Grenada844047.6945457.4Montserrat282589.3312580.1St. Kitts and Nevis1387856.51817642.0TOTAL30518059.044223452.9
Discussion {#Sec5}
==========
Although sample sizes were small, our findings provide the first evidence of *T. gondii* infections in small ruminants from Dominica, Montserrat and St. Kitts and Nevis. Our results, showing that over 40% of sheep and goats from these islands and Grenada were seropositive for *T. gondii,* are consistent with previous studies in the Caribbean that reported seroprevalences of 44% (sheep) and 43% (goats) in Grenada \[[@CR10]\], and 43% (goats) in Trinidad \[[@CR9]\]. The results are also similar to studies in the USA where seroprevalences vary from 60-74% in sheep \[[@CR7]\] and 22-65% in goats \[[@CR17]\], and in Central America where seroprevalences of 44% in goats \[[@CR18]\] and 38% in sheep \[[@CR7]\] have been reported. Similar seroprevalences have also been reported in South America \[[@CR19]\]. In our study, the highest seroprevalence was in Montserrat (over 80%); however, the sample size of animals was the lowest of the 4 island-nations and may therefore not be representative of the true seroprevalence. Also, we have no data on the age of the animals sampled, so it is possible that the animals in Montserrat were older than those sampled on the other islands and therefore more likely to be infected, since seropositivity is known to increase with age \[[@CR20], [@CR21]\].
With the exception of Grenada, *T. gondii* seroprevalence was higher in sheep than in goats which is a pattern that has previously been reported \[[@CR10], [@CR20]\] and may be a result of their foraging behaviour and diet selection. Goats are natural browsers tending to eat leaves and twigs from higher bushes and shrubs, whereas sheep are grazers tending to eat short grasses and clovers close to the soil, thus being more likely to encounter oocysts.
Previous studies in Grenada and St. Kitts and Nevis have reported a high *T. gondii* seroprevalence in cats on the islands, indicating that the environment may be contaminated with oocysts shed in their faeces. A survey of cats in Grenada indicated that 31% of domestic cats, and 28% of feral cats had antibodies to *T. gondii* \[[@CR22]\]. Studies in St. Kitts and Nevis have demonstrated that 85% of domestic cats \[[@CR23]\] and 74% of feral cats \[[@CR24]\] have antibodies to *T. gondii.* Since ingestion of oocysts from contaminated pasture, feed or water is the main route of transmission for herbivores, the high prevalence reported in the present study suggests that the farms on which the animals were sampled had cats on the premises at some point. Furthermore, previous work has demonstrated that 41% of feral chickens in Grenada have antibodies to *T. gondii*, suggesting widespread environmental contamination with oocysts, since chickens are most likely to become infected when feeding on the ground \[[@CR25]\].
It is not known whether the animals sampled in this study were destined for the local food chain; however, it is highly likely since there is no exportation of meat from these islands \[[@CR26]\] and there is usually insufficient local production of small ruminant meat to cater for local demand \[[@CR5]\]. While we cannot say whether the animals were harbouring tissue cysts, the high seroprevalence indicates high levels of exposure to the parasite, and tissue cysts are known to persist for the lifetime of the host \[[@CR2]\]. Therefore, there could be a potential public health risk should the meat be consumed raw or undercooked. In the Caribbean, it is generally common practice to overcook or pressure-cook meat which will reduce the risk of transmission of *T. gondii*; however, the high seroprevalence demonstrated in this study indicates high levels of environmental contamination with oocysts, which could potentially offer an alternative route of transmission through contaminated fruit or vegetables \[[@CR27]\] or water \[[@CR28]\].
Toxoplasmosis is a significant cause of reproductive failure in sheep and goats, leading to economic losses. Although most sheep and goat owners in the Caribbean are small-scale, low-input producers \[[@CR5]\], ruminant meat production still serves as a source of income, and any losses due to abortions may have an impact on their livelihood. There is a paucity of data on *T. gondii*-associated abortions in the Caribbean, likely because they are never reported; however, a longitudinal study on a goat herd in Tobago demonstrated increasing *T. gondii* seroprevalence which was significantly associated with increasing abortion rates \[[@CR29]\]. If there is widespread environmental contamination with oocysts on the islands in this study, as indicated by the high seroprevalence, reproductive losses due to toxoplasmosis could potentially be a source of economic losses.
**Competing interests**
The authors declare that they have no competing interests.
**Authors' contributions**
CMH performed all serological screening and data analysis, and drafted the manuscript. FK and EAI provided antigenic material and control sera for the ELISAs, and revised the manuscript. PJK conducted original studies which generated serum samples used in the present study, and also revised the manuscript. All authors read and approved the final manuscript.
This project was funded by Ross University School of Veterinary Medicine and the Moredun Research Institute. We thank the following for assistance with collection of samples: Dr R Thomas (Ministry of Agriculture, Dominica), Dr B Louison (Ministry of Agriculture, Grenada), Dr N Waldron (Ministry of Agriculture, Montserrat), and Drs T. Challenger, P Bartlett and L Henry (Ministry of Agriculture, St Kitts and Nevis).
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction
===============
The advent and continuing development of magnetic resonance imaging (MRI) techniques has allowed visualization of changes within the structure and size of the human brain in a variety of clinical conditions but, in particular, following a patient's earlier exposure to traumatic brain injury (TBI) \[[@B1-brainsci-03-01374]\]. In overview, long term changes are loss of the total volume of the brain, loss of the volume of both cerebral white and grey matter, loss in volume of the corpus callosum, thalamus, hippocampus, amygdala, some association, callosal and projection pathways \[[@B1-brainsci-03-01374],[@B2-brainsci-03-01374],[@B3-brainsci-03-01374],[@B4-brainsci-03-01374],[@B5-brainsci-03-01374]\] and an increased relative volume of the brain ventricular system ([Figure 1](#brainsci-03-01374-f001){ref-type="fig"}) have been widely reported \[[@B6-brainsci-03-01374],[@B7-brainsci-03-01374]\] up to two years after TBI \[[@B8-brainsci-03-01374]\].
![Two years prior to sustaining a severe traumatic brain injury (TBI) this patient underwent magnetic resonance imaging (MRI) as part of an assessment for persistent headache. The MRI was interpreted as within normal limits (WNL) and computation of the ventricle-to-brain ratio (VBR) was likewise WNL (average adult VBR is \~1.5 with a 0.5 standard deviation based on total ventricular volume/total brain volume multiplied by 100 so that whole numbers may be used). Note that on the DOI the anterior horns of the lateral ventricle were equally compressed being an indication of generalized cerebral edema. Likewise, the VBR was more than half of normal, a further reflection of generalized edema. By day two post injury, the VBR was further reduced, but by one week post-injury had increased with significant enlargement and was followed by cortical atrophy observed at 16 weeks post injury, with prominent further generalized cortical atrophy, ventricular enlargement and elevated VBR at two years post-injury. (Illustration from Bigler and Maxwell 2012 \[[@B1-brainsci-03-01374]\] with thanks).](brainsci-03-01374-g001){#brainsci-03-01374-f001}
The great majority of the experimental literature utilizing animal models of TBI and/or traumatic axonal injury (TAI) has reported development of pathological changes in axons of myelinated nerve fibers and their cell bodies or, to a lesser extent *in vitro* studies, rather than development of pathology within the myelin sheath of injured axons. Within the recent TAI literature, however, attention is beginning to be focused upon responses by central myelin and oligodendrocytes after TAI \[[@B9-brainsci-03-01374],[@B10-brainsci-03-01374],[@B11-brainsci-03-01374]\].
A recent major change in thinking with regard to treatment of and care of patients who have experienced a TBI is that even a mild traumatic episode may potentiate the post-traumatic development of chronic traumatic encephalopathy (CTE) where patients, over a period of months to several years after a traumatic episode, increasingly experience poor concentration, poor attention, disorientation, memory deficits, emotional and behavioral difficulties which worsen with increasing post-traumatic survival \[[@B12-brainsci-03-01374]\]. However, the pathological pathway(s) associated with development of CTE are obscure \[[@B12-brainsci-03-01374]\]. A small number of experimental studies, however, have reported that axonal degeneration may continue over months following TBI \[[@B13-brainsci-03-01374],[@B14-brainsci-03-01374]\] using markers of disrupted axonal transport or stereology. It was decided to investigate pathological changes in the glial supporting cells related to axons of central white matter in a stretch-injured optic nerve experimental, animal model to test the hypothesis that central myelin may develop post-traumatic pathology that could influence axon survival after TBI.
2. Results and Discussion
=========================
2.1. Results
------------
The earliest changes in myelin morphology may be seen at 2 h after stretch-injury in the guinea-pig optic nerve model of TAI ([Figure 2](#brainsci-03-01374-f002){ref-type="fig"}). At low magnification, numerous irregular myelin profiles occur within the myelin sheaths of large and small nerve fibers viewed in longitudinal section. In the present study these myelin profiles will be termed myelin discontinuities or (md) ([Figure 2](#brainsci-03-01374-f002){ref-type="fig"} white arrows).
At high magnification, the separation of adjacent myelin lamellae is seen to occur on the cytoplasmic surface of the oligodendrocyte plasmalemma because myelin membranes separate through the major dense line ([Figure 3](#brainsci-03-01374-f003){ref-type="fig"}) rather than the interperiod line.
A group of uninjured and injured animals were fixed and processed to localize free calcium ions by use of the modified pyroantimonate precipitation procedure \[[@B15-brainsci-03-01374],[@B16-brainsci-03-01374],[@B17-brainsci-03-01374]\]. In uninjured animals, small oval foci of separation of myelin lamellae form myelin discontinuities ([Figure 4](#brainsci-03-01374-f004){ref-type="fig"}) and within some of these aggregates of pyroantimonate precipitate mark the presence of free calcium ions ([Figure 4](#brainsci-03-01374-f004){ref-type="fig"}).
In material fixed at 1--2 h after TAI, myelin discontinuities (md) are readily apparent in both transverse and longitudinal sections ([Figure 5](#brainsci-03-01374-f005){ref-type="fig"}a,b) of nerve fibers that otherwise lack any indication that damage or injury has occurred in these nerve fibers. Pyroantimonate precipitate is localized to the md and is suggestive of a focal concentration of calcium ions within the cytoplasm within the longitudinal incisures extending through the compact myelin of internodal segments of myelin.
{#brainsci-03-01374-f002}
{#brainsci-03-01374-f003}
{#brainsci-03-01374-f004}
It is noteworthy in [Figure 5](#brainsci-03-01374-f005){ref-type="fig"}b that despite the occurrence of a diffuse pyroantimonate precipitate within the axoplasm of the left hand nerve fiber longitudinally orientated microtubules (white arrows) are numerous and intact mitochondria (mt), with no evidence of a central lucent vacuole, occur within the right hand nerve fiber. The presence of both features indicates no significant pathology within these nerve fibers, and it is suggested that the small, focal aggregates of pyroantimonate precipitate identify the cytoplasmic myelin longitudinal incisures of intact nerve fibers. The longitudinal incisures extend from the paranodal cytoplasmic loops at either end of a myelin internode ([Figure 6](#brainsci-03-01374-f006){ref-type="fig"}) \[[@B18-brainsci-03-01374]\], to form cytoplasmic channels or tunnels through the internodal compact myelin to provide for the distribution of ions and molecules throughout the myelin in an internode \[[@B18-brainsci-03-01374],[@B19-brainsci-03-01374],[@B20-brainsci-03-01374]\].
Recently, small triangular clefts between adjacent paranodal loops and the outer surface of the axolemma have been reported \[[@B18-brainsci-03-01374]\]. Solutions of ions and small molecules placed using microinjection techniques in the perinodal space pass through these spiral clefts between adjacent paranodal glial loops and the subjacent axolemma \[[@B18-brainsci-03-01374]\]. These periaxonal/oligodendrocyte plasmalemma limited spaces or clefts will here be termed "paranodal spiral clefts" and are visible in thin sections viewed at high magnification ([Figure 7](#brainsci-03-01374-f007){ref-type="fig"}a). Following stretch-injury to the optic nerve, the volume of the paranodal spiral clefts appears to increase with survival out to 24 h after injury ([Figure 7](#brainsci-03-01374-f007){ref-type="fig"}b,c). Within sections processed for calcium pyroantimonate, small aggregates of precipitate occur within the volume of the spiral clefts ([Figure 7](#brainsci-03-01374-f007){ref-type="fig"}b,c).
{#brainsci-03-01374-f005}
With increasing survival after injury there is an apparent increase in volume of myelin discontinuities (md) ([Figure 8](#brainsci-03-01374-f008){ref-type="fig"}a), together with a raised intra-axonal content of pyroantimonate precipitate. In a low magnification field at 12 h survival myelin discontinuities occur more frequently in larger nerve fibres and mds now protrude into the perimeter of the axon ([Figure 8](#brainsci-03-01374-f008){ref-type="fig"}a, arrows). However, and in particular in the context of this special issue on myelin and myelin repair, very little attention has been paid to analysis of responses within the myelin sheath of injured central white matter, for example in models of stroke \[[@B21-brainsci-03-01374]\] and in trauma \[[@B22-brainsci-03-01374],[@B23-brainsci-03-01374],[@B24-brainsci-03-01374]\].
Central myelin is often difficult to preserve for ultrastructural analysis and published experimental studies of traumatically injured myelinated nerve fibers have frequently used immunocytochemical techniques to investigate changes within the axon. There is also often use of modified microwave antigen retrieval \[[@B22-brainsci-03-01374],[@B23-brainsci-03-01374]\] or tissue homogenization \[[@B25-brainsci-03-01374]\] techniques. Damage to the myelin sheath cannot, therefore, be totally excluded during processing prior to microscopic examination. In transverse sections of stretch-injured nerve fibers at 4--12 h, mds occupy a large proportion of the cross-sectional area of the myelin sheath ([Figure 8](#brainsci-03-01374-f008){ref-type="fig"}b) being most extensive in larger axons of the optic nerve, for example compare the large central nerve fiber and the right hand smaller fiber ([Figure 8](#brainsci-03-01374-f008){ref-type="fig"}b). The normal smooth profile of the axon has been lost and the axolemma has lost its intimate relation to the inner aspect of the myelin sheath forming periaxonal spaces (pa). The axoplasm of the axon, however, contains recognizable cytoskeletal components but these have a tortuous rather than linear, longitudinal course parallel to the long axis of the axon. Mitochondria are present in the axoplasm but contain either aggregates of pyroantimonate precipitate (arrows) or a central lucent zone (double arrow). These changes within the axoplasm mirror those reported in a number of earlier studies of traumatic axonal injury \[[@B14-brainsci-03-01374],[@B16-brainsci-03-01374],[@B22-brainsci-03-01374],[@B23-brainsci-03-01374],[@B26-brainsci-03-01374],[@B27-brainsci-03-01374]\].
![Schematic view of an oligodendrocyte and the myelin sheaths maintained by that cell. On the uppermost nerve fibre, the myelin sheath is partially unwrapped. In the lowest fiber the myelin sheath is shown completely unwrapped forming a flattened sheet with only the inner loop of the oligodendrocyte process juxtaposed to the axon (labelled axon in white). Channels containing small portions of oligodendrocyte cytoplasm are indicated in orange and form the paranodal/lateral loops, and the ad (inner loop) and abaxonal (outer loop) spaces and are all in continuity with each other. Extending from the paranodal/lateral loops into the compact myelin (in green) of the sheath are four schematic longitudinal incisures. Adapted from Aggarwal *et al*. 2011 \[[@B18-brainsci-03-01374]\].](brainsci-03-01374-g006){#brainsci-03-01374-f006}
{#brainsci-03-01374-f007}
With increasing post-traumatic survival in experimental animal models of TAI, it is now recognized that axons may undergo secondary axotomy even months after injury \[[@B14-brainsci-03-01374]\]. And there is rapidly accumulating evidence that even a single incident of moderate to severe TBI is associated with progression of cognitive difficulties extending out, possibly, to dementia \[[@B28-brainsci-03-01374]\]. A number of changes occur both in the myelin sheath and the axon of such degenerating fibers. Stereological analyses have demonstrated in the guinea pig optic nerve stretch-injury model that nerve fibers of different size or caliber are lost from the optic nerve over different survival time frames. Notably, larger nerve fibers are lost rapidly, over about 4--12 h, after TAI and there is then a progressive shift to the left of the numbers of nerve fibers within 0.5 μm wide groups \[[@B14-brainsci-03-01374]\]. That is, with increasing survival over weeks and months after a single injury episode there is a more rapid loss of the larger nerve fibers within the injured optic nerve and an associated rise in the number of fibers between 1.0 and 1.5 µm in diameter although it is true that the total number of intact nerve fibers in the injured optic nerve falls over the 12 week survival of this experiment \[[@B14-brainsci-03-01374]\] when about 60% of nerve fibers in optic nerves of uninjured nerves remain. Examples of the range of pathologies within the myelin sheath of degenerating nerve fibers are provided in [Figure 9](#brainsci-03-01374-f009){ref-type="fig"}. But it should be noted that each example may be found within the optic nerve at any survival point later than 8--12 h out to 12 weeks when this experiment was terminated.
{#brainsci-03-01374-f008}
{#brainsci-03-01374-f009}
With increasing survival there is an increase in the volume of myelin discontinuities until the diameter of the locus of myelin separation become larger than the width of the myelin sheath. The expanded myelin discontinuities extend either into the lumen of the sheath and displace or injure the axon contained therein ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}a,b) or expand into the peripheral extracellular space between adjacent nerve fibers ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}c). To minimize the risk of confusion the term used to refer to intrusions into the lumen of the sheath ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}a,b) in this study shall be "myelin intrusion": for expansion of the myelin into the pericellular/extracellular space shall be "external protrusion" ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}c). That is to say that a myelin discontinuity may continue to increase in volume to form either a myelin intrusion or a external protrusion or even both when the zone of separation of adjacent myelin lamellae is especially extensive or large, as seen for example, in [Figure 8](#brainsci-03-01374-f008){ref-type="fig"}b. However, preliminary data indicates that the development of myelin intrusions may be associated with greater pathology of or damage to the axon. A separation of the surface of the axolemma and the innermost layer of the myelin sheath may be generated to form a periaxonal space ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}a,b) with related ultrastructural pathology suggestive of disruption of the axonal cytoskeleton (*cf.* [Figure 9](#brainsci-03-01374-f009){ref-type="fig"}a--c). Similar changes in the ultrastructure of the axonal cytoskeleton are a concept which have been extensively reported in the TBI literature \[[@B1-brainsci-03-01374],[@B4-brainsci-03-01374],[@B7-brainsci-03-01374],[@B26-brainsci-03-01374],[@B28-brainsci-03-01374],[@B29-brainsci-03-01374],[@B30-brainsci-03-01374],[@B31-brainsci-03-01374]\]. Two types of axonal degeneration have been reported in TBI, either so-called light/watery degeneration or a dark degeneration \[[@B32-brainsci-03-01374]\]. Examples of both light ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}e) and dark ([Figure 9](#brainsci-03-01374-f009){ref-type="fig"}f--h) degeneration may be found in sections following stretch-injury between 48 h and 8 weeks survivals.
In the guinea pig optic nerve stretch injury model of TAI classic degeneration bulbs may be encountered in any specimen from 7 days and later survivals. Two examples are illustrated in [Figure 10](#brainsci-03-01374-f010){ref-type="fig"}a,b. The specimen in [Figure 10](#brainsci-03-01374-f010){ref-type="fig"}a was obtained at 7 days post-trauma, while that in [Figure 10](#brainsci-03-01374-f010){ref-type="fig"}b was obtained at 4 weeks post-injury. In the latter, the number of rounded, darkly stained, myelin intrusions is very high within the myelin sheaths of all sizes of nerve fiber as well as in the myelin remnants in relation to the terminal degeneration bulbs. The pericellular tissue fluid internal to the remnants of the myelin sheath (top) and within the extracellular space related to the bulbs contains pyroantimonate precipitate. Mitochondria within degeneration bulbs frequently also contain small foci of precipitate. In [Figure 10](#brainsci-03-01374-f010){ref-type="fig"}a, the 7 day survival animal, the extracellular space is enlarged such that cell processes are widely separated and aggregates of membranous debris (◊) occur and is suggestive of tissue edema. A nerve fiber with the pathological characteristics of "dark degeneration" may be seen at the bottom of the figure (arrow).
![Longitudinal sections of an axonal swelling (**a**) and several degeneration bulbs (**b**). Material in (**a**) from an animal killed at 7 days after injury. The axonal swelling is formed by aggregation of large numbers of membranous organelles which accumulate at a locus of loss of fast axonal transport. The volume of axoplasm increases from the caliber of the axon (ax) and is completely enclosed by direct continuation of the internodal myelin sheath. The cytoplasm of the swelling is packed with numerous, randomly orientated mitochondria (m). At one region a constriction of the caliber of the axonal swelling occurs. This is the site at which the axonal swelling breaks the axon into two fragments when axonal disconnection occurs. The axon has then completed secondary axotomy and both fragments will die back some 600--800 µm on either side of the site of disconnection over the ensuing 48 h \[[@B23-brainsci-03-01374]\]. The surrounding tissue has widely separated cell processes and is suggestive of edema. Several aggregates of loosely organised membrane fragments (◊) occur in the enlarged extracellular space. An exrample of nerve fiber undergoing "dark degeneration" occurs at the bottom of this field. Magnification 2800×. (**b**) Longitudinal sections of terminal bulbs obtained from a 4 week posttrauma survival animal processed with the pyroantimonate technique. There is little evidence suggestive of edema. Numerous, closely spaced myelin intrusions and/or external protrusions occur within myelin sheaths of both nerve fibers or the myelin remnants adjacent to the large caliber terminal bulbs. The rounded, bulbous profile of the bulbs is enveloped by a remnant of the myelin sheath and probably represents myelin from a more proximal region of the degenerating nerve fibre during retraction of the terminal bulbs subsequent to axotomy. Mitochondria within both of the bulbs and parts of recognizable nerve fibers contain focal, electron dense aggregates of pyroantimonate precipitate---see inside the dotted ovals for example. There are also widespread aggregates of pyroantimonate precipitate at foci of extracellular fluid. Magnification 1875×.](brainsci-03-01374-g010){#brainsci-03-01374-f010}
2.2. Discussion
---------------
The present study provides novel information about changes in the organization of the myelin sheath related to injured, degenerating nerve fibers within a central white matter tract. A major conclusion is that ultrastructural evidence of an ongoing pathology in the myelin sheath occurs within a short time of mechanical insult to the white matter tract and continues throughout the entire extent of the experimental survival period. Stereological analysis of changes in the number and size of intact nerve fibers within the stretch-injured optic nerve provides evidence that loss of nerve fibers continues throughout the experimental survival of 12 weeks \[[@B14-brainsci-03-01374]\]. The present study extends those findings by reporting pathology within the myelin sheath of degenerating nerve fibers. However, a more detailed, stereological analysis is now required to improve understanding of this previously unreported pathological change.
Myelin of myelinated nerve fibers within the brain and spinal cord is formed by oligodendrocytes where a single oligodendrocyte forms and maintains the myelin sheaths of a number of separate, closely related axons where an oligodendrocyte supplies the myelin forming a single internodal segment on each fiber ([Figure 6](#brainsci-03-01374-f006){ref-type="fig"}). A circumferential, multilayered sleeve of membranes having a regular spacing at their cytosolic and external surfaces, correlated with the lipid mosaic model of biological membranes, and which can be resolved using either polarised light microscopy, or transmission electron microscopy (TEM) of either thin, resin embedded, sections or freeze-fracture replicas has been widely reported \[[@B1-brainsci-03-01374]\]. The structure observed is consistent with a protein-lipid-protein-lipid-protein organization of the oligodendrocyte's axolemma forming the myelin sheath with a higher concentration of proteins on the surface of the plasmalemma facing the cytoplasm, the so-termed cytosolic face that is stained more heavily in routinely processed resin embedding for TEM ([Figure 3](#brainsci-03-01374-f003){ref-type="fig"}). These darkly stained portions of the repeating layers of oligodendrocyte plasmamembranes are termed the "major period" or "dark period" lines of the myelin sheath ([Figure 3](#brainsci-03-01374-f003){ref-type="fig"}). That part of the plasmamembrane facing outward to the external environment around a cell contains a greater proportion of lipid and stains less darkly with routine TEM staining and forms the "interperiod line" ([Figure 3](#brainsci-03-01374-f003){ref-type="fig"}). The distance between adjacent dark period lines has been reported as 16 nm using TEM X-ray diffraction and 12 nm in resin embedded thin sections, the lower value probably reflecting the fact that tissue for resin embedding has to be markedly dehydrated \[[@B18-brainsci-03-01374],[@B33-brainsci-03-01374]\] ([Figure 3](#brainsci-03-01374-f003){ref-type="fig"}).
Until recently there was only a limited understanding of the chemical constituents of CNS myelin. But our knowledge has increased dramatically with the advent of proteomic analyses of the molecular biology of the insulating sheath of central nerve fibers. Jahn *et al*. \[[@B34-brainsci-03-01374]\] have provided information concerning nearly 350 proteins associated with central nervous system myelin and a much greater complexity of membrane associated interactions is now appreciated. Membrane components may occur in only very small relative proportions of the total content but it is also recognized that earlier estimates of relative lipid and protein content were over simplified \[[@B34-brainsci-03-01374]\]. Jahn *et al*. \[[@B34-brainsci-03-01374]\] suggested that the term "myelin-enriched" be applied to the major fraction of a white matter ultracentrifugate rather than "compact myelin" because proteins localized within the non-compacted cytoplasmic channels of the inner and outer loops, the lateral or paranodal loops and the longitudinal incisures mentioned above ([Figure 6](#brainsci-03-01374-f006){ref-type="fig"}) will occur in the ultracentrifugate.
Central, and peripheral, myelin differs from other plasma membranes in that 70%--75% of the dry weight is lipid in the proportion of 2.2 (cholesterol):1 (phospholipid):1 (galactolipid/plasmalogen) as reviewed in \[[@B34-brainsci-03-01374]\]. The largest proportions of proteins in compact myelin are proteolipid protein (PLP) (17%), myelin basic protein (MBP) (8%) and 2′,3′-cyclic nucleotide 3′-phosphodiesterase(CNP) (4%). MBP is an extrinsic membrane protein in the cytoplasmic face of compact myelin and has a high electrical charge which binds to negatively charged lipids and contributes to the integrity of the major dense line of compact myelin. In addition a variety of transmembrane proteins---myelin-associated glycoprotein (MAG) localized particularly in the periaxonal membrane and glial loops of the paranodes \[[@B35-brainsci-03-01374]\], myelin oligodendrocyte glycoprotein (MOG), tetraspanin 2, M6B, oligodendrocyte-specific protein (OSP/Claudin-11)---and cellular adhesion complexes---glial nectin-like protein (Nec14), glial neurofascin (NF155) and contactin/contactin associated protein 1 (Caspr) are important in glial-axonal junctions at the paranode. The interested reader is referred to the publication by Jahn *et al*. \[[@B34-brainsci-03-01374]\] for further information.
Recent evidence has strongly suggested that intercellular relationships, and probably functions, within central white matter are more complex than appreciated even less than a decade ago \[[@B1-brainsci-03-01374],[@B18-brainsci-03-01374],[@B19-brainsci-03-01374]\]. Rather than the earlier concept that mature oligodendrocytes assume a quiescent state upon completion of myelination, there is now a consensus that considerable intercellular exchange occurs between the neuronal axon, oligodendrocytes and perinodal astrocytes within central white matter \[[@B18-brainsci-03-01374]\]. Astrocytes and oligodendrocytes within the CNS are linked by multiple gap junctions during myelination \[[@B36-brainsci-03-01374]\] and in mature white matter \[[@B19-brainsci-03-01374]\] ([Figure 11](#brainsci-03-01374-f011){ref-type="fig"}). It has recently been reported that various types of mechanical load, like strain, pressure, shear stress, or cyclic stretch can influence oligodendrocyte cell biology and intercellular communication via gap junctions between neighboring mature oligodendrocytes as well as between mature oligodendrocytes and astrocytes \[[@B37-brainsci-03-01374]\] both of which intimately interact with CNS neurons. Gap junctions form narrow channels connecting the cytoplasm of adjacent or linked cells ([Figure 11](#brainsci-03-01374-f011){ref-type="fig"}) allowing passage of molecules or ions of less than 1000 Da and electrical current \[[@B19-brainsci-03-01374],[@B37-brainsci-03-01374]\] and such channels may be visualized at the light microscope level by use of biocytin which readily passes through gap junctions and into the oligodendrocyte cytoplasm \[[@B19-brainsci-03-01374]\] or at the ultrastructural level ([Figure 11](#brainsci-03-01374-f011){ref-type="fig"}).
{#brainsci-03-01374-f011}
Studies have also demonstrated that ions and small molecules may pass from the extracellular space, for example from the perinodal space, into the lateral cytoplasmic loops at the paranodes, then into the longitudinal incisures passing through the internodal compact myelin as represented in [Figure 6](#brainsci-03-01374-f006){ref-type="fig"}. Oligodendrocytes are now recognized to allow exchange of metabolites \[[@B38-brainsci-03-01374]\], ions and other gap-junction permeable molecules between neighboring cells and this is currently thought to have a major influence during myelination \[[@B18-brainsci-03-01374]\].
There is a consensus that following TBI a wave of calcium mediated depolarisations spread from the locus of injury via interconnected gap junctions between astrocytes \[[@B19-brainsci-03-01374],[@B39-brainsci-03-01374]\] and allow propagation of astrocyte Ca^2+^ waves over distances of several hundred micrometers from an initial locus \[[@B19-brainsci-03-01374],[@B38-brainsci-03-01374],[@B40-brainsci-03-01374],[@B41-brainsci-03-01374]\]. This leads to the postulate that following TAI, bi-directional Ca^2+^ waves occur in electrically interconnected astrocytes and oligodendrocytes in central white matter. The maintenance of abnormal intracellular Ca^2+^ reported *in vitro* \[[@B42-brainsci-03-01374]\] over at least 24 h after stretch injury provides support for this hypothesis. It is suggested that the present study provides further support for the hypothesis that glia \[[@B43-brainsci-03-01374]\], both astrocytes and oligodendrocytes, respond to axonal injury by allowing Ca^2+^ waves to propagate over hundreds of micrometers through gap junctions. The findings in the present study allows suggestion that Ca^2+^ waves or "spreading depression" serve to exacerbate the number of injured nerve fibers through metabolic compromise as a result of damage to myelin.
Nawaz *et al*. \[[@B44-brainsci-03-01374]\] reported that an increasing intracellular Ca^2+^ level resulted in a rapid dissociation of MBP from the plasma membrane through a phospholipase C (PLC) dependent hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP2). It was hypothesized \[[@B44-brainsci-03-01374]\] that abnormal Ca^2+^ entry and Ca^2+^ dependent myelin delamination in white matter tracts is likely to be caused by the detachment of MBP from myelin membranes. Following stretch injury to optic nerve fibers of the guinea pig, it is posited that Ca^2+^ depolarization via gap junctions linking oligodendrocytes and astrocytes results in elevated levels of Ca^2+^ throughout the longitudinal incisures of internodal myelin segments and results in focal activation of intrinsic calpains which have been reported to be associated with proteolysis of MBP following TBI \[[@B9-brainsci-03-01374]\]. Delamination of myelin, in association with altered metabolism in myelin, oligodendroctytes and perinodal astrocytes potentiates loss of regulation of axonal calcium homeostasis and allows the potential recruitment of axons to undergo secondary axotomy. The provision of quantitative data in support for this hypothesis is currently being investigated and the established clinical and psychological paradigms that suggest an on-going loss of white matter with increasing post-traumatic survival provides a stimulus to further experimental study. These investigations may provide a therapy to minimize loss of white matter with increasing post-traumatic survival rather than, at present, providing increasing numbers of patients, and their families, the opportunity to experience the long-term loss of neurological function that appears to be the current outcome after TBI.
3. Experimental Section
=======================
Under intramuscular ketamine (50 mg kg^−1^) and xylazine (3 mg kg^−1^) anaesthesia, the right optic nerve of adult guinea-pigs (range 700--850 g) was stretched, in a controlled manner \[[@B25-brainsci-03-01374]\], to provide reproducible and measurable amounts of elongation or tensile strain \[[@B16-brainsci-03-01374]\]. Animals were fixed either by transcardiac perfusion with 2.5% gluteraldehyde in 0.2 M PIPES buffer or fixed for the ultrastructural demonstration of the mobile pool of Ca^2+^ (Borgers *et al*., 1977 \[[@B15-brainsci-03-01374]\]) with the slight modification that 0.5% paraformaldehyde and 1.9% sucrose was added to the fixative. For routine, thin section examination blocks of injured (right optic nerve) and internal control (left optic nerve), three animals at each time point (15 min, 1, 2, 4, 24, 48 h and 7, 14, 28 and 84 days and three controls) were post-fixed in 1% osmium tetroxide in PIPES buffer, through graded concentrations of alcohol, alcohol and epoxypropane, epoxypropane and embedded in Araldite using epoxypropane as an intermediary. For demonstration of the mobile pool of Ca^2+^, three animals at each time point (15 min, 1, 2, 4, 24, 48 h and 7, 14, 28 and 84 days and three controls) were terminally anaesthetized with IP barbiturate and, after thoracotomy, perfusion fixed through the left ventricle with \~90 nM potassium oxalate in 1.9% sucrose, adjusted to pH 7.4 with potassium hydroxide, at 37 °C, followed by 3% glutaraldehyde, 0.5% paraformaldehyde, 90 nM potassium oxalate, 1.9% sucrose (750--850 mosmol) adjusted to pH 7.4 with potassium hydroxide for 1 h. The first 500 mL of the fixative was warmed to 37 °C and perfused rapidly in contrast to the remaining 2000ml which was cooled to approximately 4 °C and perfused more slowly. After perfusion, both optic nerves (the left nerve used as an internal control) were dissected out and each divided into three equal segments which were subsequently processed separately. The segments were placed in the same fixative at 4 °C for 2 h, briefly rinsed in 90 mM potassium oxalate in 1.9% sucrose (pH 7.4) and postfixed in 1% osmium tetroxide and 2% potassium pyroantimonate for 2 h at room temperature. Unreacted pyroantimonate was washed out with distilled water adjusted to pH 10 with potassium hydroxide for 15 min. The segments were then routinely dehydrated through a series of 50%, 70%, 90% and three times 100% ethanols before being cleared in two 20 min changes of propylene oxide. Finally, the segments were placed successively in 1:1 propylene oxide/Araldite, 1:2 propylene oxide/Aratdite, two 4 h changes of pure Aratdite before embedding in pure Araldite and polymerized at 60 °C for 24 h. Semithin and thin sections were cut on an ultramicrotome and examined, unstained, in a Phillips 300 TEM. For routine TEM, ultrathin sections were stained with 12.5% methanolic uranyl acetate and lead citrate.
4. Conclusions
==============
Transmission electron microscopy of stretch-injured central nerve fibers provides morphological evidence which supports the hypothesis that injury to central myelin may contribute to continued axonal degeneration following traumatic brain injury. Foci of disassociation of lamellae within the myelin sheath, here termed myelin discontinuities occur in both uninjured and injured nerve fibers. However, within injured nerve fibers the spatial extent or volume of delamination of myelin lamellae increases with post-traumatic survival after stretch-injury. In parallel, the volume of paranodal spiral clefts which communicate with myelin channels/discontinuities increases with increasing post-traumatic survival. Pyroantimonate studies indicate that these foci of delamination contain increased content of free calcium. It is suggested that elevated levels of calcium activate calpains which disrupt MBP floatage of leaflets of the myelin sheath. This allows generation of the hypothesis that waves of calcium depolarization potentiate areas of myelin delamination in distant nerve fibers and the associated rise in content of free calcium compromises axonal physiology leading to recruitment of additional nerve fibers and greater damage to central nerve fiber pathways.
The author declares no conflict of interest.
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .405}
{#sp2 .406}
{#sp3 .407}
{#sp4 .408}
{#sp5 .409}
{#sp6 .410}
{#sp7 .411}
{#sp8 .412}
{#sp9 .413}
{#sp10 .414}
{#sp11 .415}
{#sp12 .416}
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#s0005}
===============
Chronic and degenerative diseases in middle-aged women such as cancer, cardiovascular, and metabolic diseases are increasing. These diseases are the leading cause of death and disability disorders in developing countries. [@b0075] showed that large changes in behavior patterns due to the decreasing consumption of vegetables and fruits, which plays an important role in increasing the incidence of chronic and degenerative diseases. Bioactive compounds in fruits and vegetables such as polyphenols, flavonoids, anthocyanins, micronutrients, minerals and vitamins, have antioxidant and anti-inflammatory activities ([@b0140]). Therefore, they possess preventive and therapeutic potencies against diseases ([@b0010], [@b0160]).
Date palm (*Phoenix dactilyfera* L.) seed is one of the rich source of polyphenols and flavonoids ([@b0035]). It has been extensively investigated for pharmacological activities such as anti-inflammatory ([@b0115]), immuno-stimulant ([@b0105]), antidiabetic ([@b0045]), antibacterial ([@b0025]), antiviral ([@b0060]), and antioxidant ([@b0020], [@b0035], [@b0050], [@b0090]). In previous studies, the date palm seeds were proven to work as anti-inflammatory ([@b0120], [@b0125], [@b0120]), and antiatherogenic food substance ([@b0110]). Recent study showed that various compounds act as anti-inflammatory by metabolomic approach ([@b0005]). Anti-inflammatory effect of date seeds in human has not been adequately studied, especially in middle-aged women. This study, therefore, aims to examine the anti-inflammatory mechanisms of steeped date palm seed in middle-aged women.
2. Material and methods {#s0010}
=======================
2.1. Study design and participants {#s0015}
----------------------------------
This is a quasi-experimental design with a pre and post-test approach. Data were obtained from a total of 30 randomly selected middle-aged women living at Gununglurah, Cilongok District, Banyumas. The inclusion criteria were women between the age of 45--60 years, with no history of metabolic disease by measuring their blood glucose and total cholesterol levels. These women live in farming community and perform the domestic tasks of housewives. In this study, each subject received a single dose of 2.5 g seed powder per day for 14 days. The seed powder was consumed using 250 mL of boiling water. Subjects signed an informed consent before participating in this study, which was conducted after obtaining approval of ethical clearance from the medical research ethics committee of the Faculty of Medicine University of Negeri Sebelas Maret Surakarta, number: 541/IV/HREC/2019.
2.2. Preparation of date seeds powder {#s0020}
-------------------------------------
Deglet Nour dates collected from Tunisia, were obtained from a market at Purwokerto, Banyumas Regency, Indonesia. Seeds were manually separated from the date flesh, cleaned with water, selected, and dried for one day in sunlight. The dried seeds were roasted at medium temperature, crushed in a blender and filtered to obtain a fine powder.
2.3. Measurement pro-inflammatory mediators {#s0025}
-------------------------------------------
A 3 mL blood sample was taken through the median cubital vein before and after treatment. The level of IL-1β, TNF-α, IL-6, IL-12, TGF-β, COX-1, COX-2, and prostaglandin E2 were examined by ELISA kit (BT Laboratories, Shanghai) based on the manufacturer\'s protocol, using an ELISA machine Reader (Labotrone, Germany).
2.4. Statistical analysis {#s0030}
-------------------------
The mean score was compared before and after treatment with all data presented as mean ± SEM. Differences in scores before and after treatment were analyzed using the student *t*-test. The graph pad software (GraphPad Prism, San Diego, CA) was used to analyze statistical and graphical data, with a p value \< 0.05 considered significant.
3. Results {#s0035}
==========
3.1. Characteristics of participants {#s0040}
------------------------------------
Thirty subjects were initially recruited, however, eight were unable to complete the treatment. Therefore, a total of 22 subjects were included in the statistical analysis. The age of subjects ranged was from 51 to 57 years, with a healthy weight status and an average body mass index of 24 kg/m^2^, blood glucose below 140 mg/dL, and total cholesterol below 200 mg/dL ([Table 1](#t0005){ref-type="table"}).Table 1Respondent characteristic.VariableRangesMeanAge (years)51-5755Blood pressure Systole (mmHg)100-175135 Diastole (mmHg)65-10093Body mass index (kg/m^2^)23.5--24.524
3.2. Expression of proinflammatory cytokines {#s0045}
--------------------------------------------
The expression of IL-1β and TGF-β reduced significantly after the consumption of date palm seeds in the post-treatment stage, while TNF-α, IL-12, and IL-6 did had no significant change ([Fig. 1](#f0005){ref-type="fig"}). This result suggests the date palm seed reduces IL-1β and TGF-β produced in healthy middle-aged women.Fig. 1**Treatment date palm seed decreases IL-1β and TGF-β expression.** The cytokines levels were measured at day 0 (pre-treatment) and day 15 (post-treatment) by enzyme-linked immunosorbent assay (ELISA). Mean ± SE are presented (n = 22). Asterisks indicate student *t*-test significance values; \*\*P \< 0.01, \*P \< 0.05. ns = not significant.
3.3. The role of dates palm seed on cyclooxygenase pathway {#s0050}
----------------------------------------------------------
The cyclooxygenase (COX) pathway is important in arachidonic acid metabolism related to the inflammatory process ([@b0080]). Therefore, the expression of COX-1, COX-2, as well as prostaglandin E2 level, showed that these mediators are important markers on the COX pathway. A significant decrease in the level of COX-1, COX-2 and PGE2 was observed in the post-treatment stage compared with pre-treatment of steeped of date palm seeds ([Fig. 2](#f0010){ref-type="fig"}). Interestingly, the expression of COX-2 substantially decreased (p \< 0.001), which indicates that the inducible expression of COX-2 in middle-aged women was decreased by consumption of date palm seeds.Fig. 2**Treatment date palm seed decreases COX-1, COX-2, PGE2 level.** The enzyme/mediator levels were measured at day 0 (pre-treatment) and day 15 (post-treatment) by enzyme-linked immunosorbent assay (ELISA). Mean ± SE are shown (n = 22). Asterisks indicate student *t*-test significance values; \*\*P \< 0.01, \*\*\*\*P \< 0.001.
4. Discussion {#s0055}
=============
Several studies have been conducted to examine the potential of date palm seeds as an anti-inflammatory ([@b0015]) *in vitro* and *in vivo* ([@b0035], [@b0085], [@b0115]). However, no anti-inflammatory activity studies have been conducted on date palm seeds in humans, women middle age. This study, found that the expression of IL-1β, TGF-β, COX-1 and COX-2 decreased after the administration of date palm seeds to middle age women.
Studies showed that the elderly and middle age women experience an increase in free radicals due to various factors such as aging, food, pollution, and excess activity. Free radicals in the body causes oxidation in normal cells, thereby, leading to inflammation and diseases ([@b0135], [@b0145]). Lymphocytes and other immune cells produce cytokines such as IL-1β and TGF-β when there are inflammatory stimuli.
Interleukin-1β is highly elevated in chronic diseases such as obesity, osteoarthritis, and gout ([@b0030]). IL-1β affects lymphocytes and macrophages, induces the formation of prostaglandins, colony stimulating factors and other cytokines ([@b0165]). IL-1β expression is stimulated by various stimuli including sterile stimulus through the formation of inflammasomes ([@b0130]). Since middle age women have developed accumulated agents capable of activating the inflammasome formation, such as cholesterol and uric acid ([@b0095]), it is therefore, important to determine the ability of the date palm seed to reduce the expression of IL-1β. Another interesting aspect is to reveal the role of date palm seeds in inhibiting IL-1β expression, which is further investigated in future.
Although TGF-β was initially called as anti-inflammatory cytokines, it also induces inflammatory Th17 cells differentiation ([@b0170]). Although the role TGF-β in inflammation remains unclear, the date palm seed showed its ability to decrease.
Cyclooxygenase pathway has been well-established for their role in inflammation ([@b0055]). Conversion of phospholipids to arachidonic acid is mediated by the enzyme cyclooxygenase. COX-1 is constitutively expressed and plays a role in the protection of the gastric mucosa. COX-2 is expressed inducibly by an inflammatory stimulus ([@b0155]). Furthermore, various stimuli tend to induce COX-2 expression, which tends to occur with increasing age. In this study, COX-2 expression greatly decreased after the administration of date palm seeds to middle age women. Both enzymes induce the conversion of arachidonic acid into prostaglandins, e.g. PGE2 which plays a role in the vasodilation of blood vessels and increased vascular endothelial permeability ([@b0070]). The expression of both COX enzymes is inhibited by date palm seeds, with a decrease in PGE2 production.
This study is limited to high of dropping out because the subject did not complete the consumption of date palm seed powder as assigned. In addition, the food consumed by the subjects and likely to affect the level of proinflammatory mediators in the human body were not recorded.
Date palm seeds may work as an anti-inflammatory and improve the performance of the immune system ([@b0105]). [@b0100] stated that it works to suppress NF-κB, COX-1 and COX-2 enzymes, thereby, decreasing pro-inflammatory mediators. The anti-inflammatory activity of the aqueous extract of date palm seeds is related to components of polyphenols such as caffeoyl hexoside, 5-O-caffeoyl shikimic acid isomers, hydrocaffeic acid, and isorhamnetin ([@b0150], [@b0065]). This ingredient has also been proven safe for liver and kidney ([@b0040]), therefore, it can be consumed regularly by middle age women to maintain health status, improve immune systems, and prevent chronic diseases. Clinical implications obtained from this study shows that the physicians may suggest the steeped of palm seed powder as a functional beverage.
5. Conclusion {#s0060}
=============
Dates seeds can act as an anti-inflammatory by reducing interleukin-1β, TGF-β, cyclooxygenase-1 and -2 expression. Regular consumption of date palm seeds increases the body\'s immunity and prevent chronic diseases.
Declaration of Competing Interest
=================================
The authors declared that there is no conflict of interest.
{#s0080}
**Availability of Data and Materials**
All data generated or analyzed during this study are included in this manuscript. Raw data are available from the corresponding author on a reasonable request.
Acknowledgments {#s0065}
===============
The authors are grateful to Dr. Hernayanti for her laboratory support; to Lita Kusuma and Nina Setiowati for critically reviewing the manuscript.
Funding {#s0070}
=======
This study was supported by The Ministry of Research Technology and Higher Education of Indonesia through PDUPT funding.
Authors' contributions {#s0075}
======================
SY, W, AI conceived and designed the experiments; SY, W, AI performed the experiments; SM analyzed the data; SY, W, AI contributed chemicals/reagents/materials/analysis tools; SY and SM wrote the paper. The authors read and approved the final manuscript.
Peer review under responsibility of King Saud University.
| {
"pile_set_name": "PubMed Central"
} |
Dengue is an arthropod-borne febrile disease caused by a flavivirus with four serotypes (DEN-1, 2, 3 and 4) which causes an estimated 50 million infections each year[@b1]. In humans, immunity against a particular serotype is considered permanent after the exposure and cross immunity to other serotypes is considered short lived[@b2] although some studies argue for a longer duration of cross-immunity[@b3][@b4]. As a consequence, the proportions of viral serotypes co-circulating at any point in time are strongly dependent on previous incidence patterns of the disease, which determine the number of individuals susceptible to each serotype at any point in time.
Dengue transmission is also modulated by environmental conditions, among which, temperature, due to its effects on the vector reproduction, stands out as a strong predictor of incidence[@b5][@b6]. In places with sufficient seasonal temperature variation, dengue is predominantly a summer disease. So it is fair to say that these environmental fluctuations play a key role in determining beginning and end of epidemic periods. This climatic influence is exerted mainly through its effects on the force of infection, which cannot be taken as constant[@b7] but rather as a seasonal (oscillating) function of time. The long term dynamics of dengue is also modulated by the alternation of virus types in circulation. Demographics also plays a role in replenishing the population of susceptibles.
The attack ratio (AR) of a disease is a measure of morbidity defined as the number of new cases divided by the population at risk. For dengue epidemics, it can be difficult to calculate the AR due to the lack of knowledge of the population at risk. The population at risk in this case is the number of susceptibles to the circulating virus type(s) before a given epidemic. Thus, in order to calculate the attack ratio, we need to determine the number of susceptibles to the circulating virus types right before the epidemic, which is virtually impossible without regular virological surveys.
The attack ratio is also influenced by the reproductive number of the disease[@b8][@b9], which is closely associated with the force of infection. Thus the incorporation of the effective reproduction number, *R*~*t*~, as a function of time, is crucial to an accurate estimation of the AR of seasonal diseases like dengue and Influenza.
Other methods for estimating the number of susceptibles while accommodating time-varying force of infection have been proposed before, for measles[@b10][@b11], a disease that shows remarkable seasonality. These methods try to reconstruct the entire series of infectious and susceptibles from case data using deterministic models and generally work well for measles because there is a one-to-one relationship between exposure and immunity, since measles is caused by a single-strain pathogen. Recently, methods in the same fashion were developed for dengue when serotype-specific data is available[@b12]. When such data is not available, the series of susceptibles to all possible serotypes, cannot be reconstructed based solely on a deterministic transmission model, since the arrival/re-emergence of new serotypes, an intrinsically stochastic event, can drastically change the pool of susceptibles, throwing off any sequential estimation based on the incidence dynamics.
Dengue has been reintroduced in Rio de Janeiro in 1986 after being absent for 68 years[@b13]. During the last decades of the 20th century only DEN-1 and DEN-2 serotypes were in circulation. The remaining serotypes DEN-3 and DEN-4, arrived respectively in 2000 and 2011[@b14][@b15]. Due to this patterns of recent re-introduction of the disease, its incidence dynamics is still dominated by the introduction events and environmental determinants of transmission.
In this paper, we propose a new approach to estimate the number (fraction) of susceptibles using a simplified model of dengue transmission based on a single-strain Susceptible-Infectious-Removed (SIR) model with time-varying infection rate. In order to bypass the limitations of not knowing the serotype-specific seroprevalence and the exact behaviour of the force of infection through time, we propose to inform the time-varying transmissibility using the *R*~*t*~ series derived from the notification data[@b16]. We extend a Bayesian framework previously used to estimate the number of susceptibles in Influenza epidemics in Europe[@b17] to include time-varying force of infection and derive a probability distribution for *R*~*t*~ to accommodate uncertainty in the estimates. Then, from the incidence series and the population at risk, we calculate the attack ratio for each epidemic. We apply our method to estimate *S*~0~ before every major dengue epidemic in the city of Rio de Janeiro, Brazil in the last 18 years.
Methods
=======
In this section we will start by describing the data and then the method used to estimate the effective reproductive number, *R*~*t*~, from the data and obtain its posterior distribution. We then proceed to describe the Susceptible-Infectious-Recovered (SIR) model used to represent the aggregated disease incidence and how *R*~*t*~ can be integrated into the model to allow for time varying force of infection. Next, an approach to approximate the posterior distributions of the numbers of susceptible to the main circulating dengue viruses for each epidemic is detailed. Finally, we discuss how to estimate the attack ratio of each epidemic using the estimated susceptible fraction and the observed incidence.
Data
----
The data used in this paper consists of time series of weekly notified cases of dengue for the city of Rio de Janeiro from 1996 to 2014. The cases are notified based only on clinical criteria. Laboratory confirmation and serotype information are available only for a very small sample and only on recent years (2010--2013). For the parameter estimation procedures incidence was normalized by dividing the number of cases reported by the total city population at each year as given by the census (Census Bureau, Brazilian Institute of Geography and Statistics, <http://www.ibge.gov.br/english/>).
Estimating the effective reproductive number (*R* ~*t*~)
--------------------------------------------------------
In monitoring of infectious diseases, it is important to assess whether the incidence of a particular disease is increasing significantly, in order to decide to take preventive measures. The effective reproductive number at time *t*, *R*~*t*~, can be understood as a real-time estimate of the basic reproductive number (*R*~0~) and is defined as the average number of secondary cases per primary case at time *t*.
Let *Y*~*t*~ be the number of reported disease cases for a particular time *t* ∈ (0, *T*). Nishiura el al. (2010)[@b16] extend the theory developed by Stallybrass *et al.* (1931)[@b18] and propose to estimate *R*~*t*~ as
where *n* is taken to be the ratio between the length of reporting interval and the mean generation time of the disease. Here we are interested in the simpler case *n* = 1. If *R*~*t*~ is to be used as a decision tool, however, one needs to be able to quantify the uncertainty about estimate in equation [1](#eq1){ref-type="disp-formula"}. Here we detail how to obtain credibility intervals for *R*~*t*~ under the assumption that the counts *Y*~*t*~ are Poisson distributed for all *t*.
We explore the approach of Ederer and Mantel[@b19], whose objective was to obtain confidence intervals for the ratio of two Poisson counts. Let *Y*~*t*~ \~ *Poisson*(*λ*~*t*~) and *Y*~*t*+1~ \~ *Poisson*(*λ*~*t*+1~) and define *W* = *Y*~*t*~ + *Y*~*t*+1~. The authors note that by conditioning on the sum *W*
Let be such that . Analogously, define such that . Ederer and Mantel (1974)[@b19] show that one can construct a 100*α*% confidence interval for *R*~*t*~ by noting that
Because the transform from *θ* to *R*~*t*~ is monotonically increasing, the result holds for confidence and credibility intervals alike.
Many authors have chosen to quantify the uncertainty about *θ* following orthodox approaches (see for example[@b20][@b21]) mainly for simplicity. We choose instead to take a Bayesian approach and use the 100*α*% posterior credibility interval for *θ*~*t*~ as *c*~*α*~(*θ*~*t*~). If we choose the conjugate beta prior with parameters *a*~0~ and *b*~0~ for the binomial likelihood in (2), the posterior distribution for *θ*~*t*~ is
Combining equations [(4](#eq12){ref-type="disp-formula"}) and ([5](#eq13){ref-type="disp-formula"}) tells us that the induced posterior distribution of *R*~*t*~ is a beta prime (or inverted beta) with parameters *a*~1~ = *Y*~*t*+1~ + *a*~0~ and *b*~1~ = *Y*~*t*~ + *b*~0~[@b22]. The density of the induced distribution is then
Thus, the expectation of *R*~*t*~ is *a*~1~/(*b*~1~ − 1) and its variance is *a*~1~(*a*~1~ + *b*~1~ − 1)/((*b*~1~ − 2)(*b*~1~ − 1)^2^). Note that this result holds only for *n* = 1. Sampling from the posterior in (6) can be made straightforward by first sampling from (5) and then applying the transform in (4). Also, one can choose *a*~0~ and *b*~0~ so as to elicit meaningful prior distributions for *R*~*t*~. We show how to elicit the prior for *R*~*t*~ from specified prior mean and variance or coefficient of variation in the [Appendix](#S1){ref-type="supplementary-material"}.
Also, since *R*~*t*~ \> 1 indicates sustained transmission, one may be interested in computing the probability of this event. This can be easily achieved by integrating (6) over the appropriate interval. By noting that
one can compute the desired probability while avoiding dealing with the density in (6) directly.
Mathematical modelling
----------------------
A Susceptible-Infectious-Removed (SIR) model is proposed to model dengue dynamics. In the traditional formulation of the model, transmission is governed by a constant transmission rate *β* and recovery happens at a rate *τ*.
For our analysis we chose to let the force of infection vary with time, just as it does in the actual epidemics, as seen in the data. So as the epidemic progresses, the effective transmission rate changes and is given by
where *R*~*t*~ is the effective reproductive number, estimated as in 1 and *S*~0~ is the initial fraction of susceptible individuals. The complete model with the time-varying force of infection is given by the system of ordinary differential equations:
where *S* + *I* + *R* = 1 ∀ *t*. Of course, this is a rather simplified model, in which, for instance, the vector is omitted. The rationale for this simplification is based on the ability of the empirically derived *R*~*t*~ to incorporate the effects of the fluctuating vector populations. Although demography is not included in the model -- population is assumed to remain constant throughout an epidemic -- the population variation from year to year is taken into consideration as the prevalences are calculated for each year by dividing the number of cases by the official population estimated by the census. In Brazil, dengue affects all age groups which still haven't been exposed to all 4 serotypes. This is in contrast with countries in which the four serotypes have been endemic for a very long period of time, where dengue mostly affect the youth.
Also, although there are multiple circulating serotypes, our approach can not discriminate between them due to the lack of serotype-specific data. Nevertheless, this modelling strategy can still provide some insight into the disease dynamics and allows us to estimate the initial fraction of susceptibles *S*~0~, a key epidemiological parameter.
Bayesian parameter estimation
-----------------------------
We take a Bayesian approach to the estimation of *S*~0~. First the incidence time series was divided into *J* = 13 epidemic windows that corresponded to significant raises in incidence and normalized to lie on the \[0, 1\] interval. For a given interval we observe an incidence time series **Y**~**j**~. We are thus interested in the posterior distribution
The likelihood *l*(**Y**~**j**~\| ⋅) is a approximated by a Normal distribution with fixed variance *σ*^2^. This approximation is a numerical convenience since it confers better stability to the MCMC sampling. In this estimation procedure we kept *R*~*t*~ fixed at fixed at the posterior mean obtained as described above and fixed *τ* = 1/7days^−1^. To complete the inference, we need to specify prior distributions for the parameters of interest. We place a flat Beta(1, 1) prior on *S*~0*j*~ ∀ *j*
To approximate the posterior in (11) we use Markov chain Monte Carlo techniques implemented in the Bayesian inference with Python (BIP)[@b17] available at <http://code.google.com/p/bayesian-inference/>. BIP uses a Differential Evolution Adaptive Metropolis (DREAM)[@b23] scheme that efficiently samples from high-dimensional joint distributions using multiple adaptive chains running in parallel with delayed rejection (see [Appendix](#S1){ref-type="supplementary-material"} for details). Also, as the numerical integration routine implemented within BIP needs *β*(*t*) to be available at arbitrary values of *t*, i.e., as continuous function of time because of the variable step size, we used linear interpolation to obtain values of *R*~*t*~ for any time point. In this study we used one chain per parameter, i.e, 3 chains for each run. The chains were run until 5000 samples were obtained after discarding 500 burn-in samples. Convergence of the parallel chains was verified at every 100 iterations by the calculation of the Gelman-Rubin's R (potential scale reduction factor), which approaches 1 at convergence[@b24].
Calculating the attack ratio
----------------------------
The attack ratio of an epidemic is defined by the number of infections divided by the size of the population at risk.
Based on what has been discussed so far, we can rewrite (12) for each epidemic *j* as
where *S*~0*j*~ is the number of susceptibles before each epidemic *j*, which we estimated before.
Python and R code to perform all the analyses described above is publicly available at <https://github.com/fccoelho/paperLM1>.
Results and Discussion
======================
In this paper we propose a method to bypass the lack of serotype-specific case data by informing the time-varying force of infection with the instantaneous reproductive number, *R*~*t*~ which we calculate from aggregated data. The main contribution of this paper can be summarized in the following items: (i) we develop a method to quantify uncertainty about *R*~*t*~ that is readily applicable to other diseases and; (ii) *R*~*t*~ is used to inform a dynamic epidemic model with time-varying force of infection in order to gain insight into the attack ratio of each epidemic; (iii) we propose an estimation procedure for circulating serotype's *S*~0~ from aggregate case data, which is robust to epidemic sizes; (iv) AR estimates are provided for 18 years of Dengue epidemics in Rio de Janeiro, Brazil.
[Figure 1](#f1){ref-type="fig"} shows the *R*~*t*~ series, according to Equation [1](#eq1){ref-type="disp-formula"} [@b16] along with the confidence bands derived in this paper. It can be seen that the inter-epidemic periods are characterized by *R*~*t*~ being indistinguishable from 1. Due to the intrinsic variability of the *R*~*t*~ series, the examination of its credible intervals is essential to identify periods of sustained transmission. The wider intervals between epidemics are due to the scarcity of cases during these periods. credibility intervals, and therefore offers protection against false alarms (see the section on tail behaviour in the [Appendix](#S1){ref-type="supplementary-material"} for a detailed explanation).
A key epidemiological quantity is the attack ratio (AR) of an epidemic, a measure of morbidity and speed of spread which can be used to predict epidemic size and help efficient Public Health planning. The AR depends fundamentally on the population at risk, which in the case of dengue is every naive (to a particular serotype) individual in the population. Estimating the initial susceptible fraction *S*~0~ for each epidemic is thus central to the estimation of the AR. Methods for estimating the number of susceptibles have been proposed before, for other diseases[@b10][@b11]. These methods attempt to reconstruct the entire series of infectious and susceptibles for measles outbreaks from case data. In the case of dengue, the full (multi-year/multi-epidemic) series of susceptibles to all possible serotypes, cannot be reconstructed based solely on a deterministic transmission model, since the arrival/re-emergence of new serotypes (which are a stochastic events) can change drastically the pool of susceptibles throwing off any sequential estimation based on the incidence dynamics.
Since there is very limited information regarding the actual proportions of each virus in circulation and most information available is about the predominant serotypes for some epidemics in the period of study only[@b25], we propose the use of a simplified a single strain model. The main argument we put forward is that by conditioning on the *R*~*t*~ series, we implicitly take into account the variability introduced by the co-circulation of multiple serotypes and heterogeneous levels of immunity in the general population. We sought to deal with all important sources of uncertainty impinging on the estimation of the AR of a dengue epidemic, but not all could be satisfactorily addressed in this analysis. For instance, in any given epidemic there is a large number of mild and asymptomatic cases, which nevertheless acquire immunity. It is estimated that for every case reported, up to 10--20 are not seen by health authorities[@b26]. Another source of uncertainty is under-reporting of diagnosed cases, which is a serious issue in the health care systems of many developing countries such as Brazil. Duarte and França (2006)[@b27], estimated the sensitivity of Dengue reporting for hospitalized patients in Belo-Horizonte, Brazil to be of 63%, meaning that approximately 37% of the suspected Dengue cases go unreported. Lastly, demography and migrations affect the number of susceptible in ways which are not easy to fully determine.
[Figure 2](#f2){ref-type="fig"}, shows the model from (10) fitted to the data. Despite its limitations, our simplified model fits the data well. In it we can see that the susceptibles series in each epidemic starts at the estimated level of *S*~0~. The proportion of susceptibles may seem low, but we must remember that these estimates are being affected by an unknown under-reporting factor, which experts suggest is somewhere between 5 and 10, i.e. for every case observed there are 5 or 10 unobserved. Since this under-reporting affects both the numerator and denominator of (13), its effects should cancel out, giving us an unbiased attack ratio estimate. One other possible source of bias which would lead to the underestimation of *S*~0~ could come from a significant part of the population not being exposed to the disease. However, as we can see in [Fig. 3](#f3){ref-type="fig"}, despite the differences in intensity (incidence), the entire city seems to be at risk, with no particularly "protected" areas, at least in the last four epidemics.
[Table 1](#t1){ref-type="table"} contains the attack ratios and medians of the *S*~0~ estimated for each epidemic/outbreak. Underreporting of cases, which is known to exist but of which exact figures cannot be determined, will lead to underestimation *S*~0~. However, the attack ratio shall remain unbiased as the underreporting affects both the numerator and the denominator of equation [13](#eq18){ref-type="disp-formula"}. It is interesting to notice that the larger epidemics, in terms of peak size are not the one with the greater attack ratios. This stresses the importance of knowing the immunological structure of the population. Knowing the *S*~0~ for the circulating viruses we can order to more accurately assess the potential impact of a coming epidemic, since particularly virulent types, can be rendered less of a threat by a low *S*~0~. Honório *et al.*[@b28] conducted a serological survey in three separate localities within the city, right before the 2008 epidemic, the authors report seroprevalences varying from 56--77.4% which is compatible with our prediction of 87.5% (1 − *S*~0,2007~) for the entire city, considering that we underestimate *S*~0~ due to the underreporting of cases.
We hope that the results presented in paper will motivate public health authorities to invest in annual serological surveys, to determine the susceptibility profile to each dengue virus as well as to estimate the under-reporting factor of the notification system.
Additional Information
======================
**How to cite this article**: Coelho, F. C. and Carvalho, L. M. d. Estimating the Attack Ratio of Dengue Epidemics under Time-varying Force of Infection using Aggregated Notification Data. *Sci. Rep.* **5**, 18455; doi: 10.1038/srep18455 (2015).
Supplementary Material {#S1}
======================
###### Supplementary Information
LMC is grateful to Dr. Leonardo Bastos for useful discussions on the posterior inference for *R*~*t*~. The authors are also grateful to Claudia T. Codeço for helpful discussions about the manuscript.
**Author Contributions** F.C.C. and L.M.C. wrote the main manuscript. L.M.C. and F.C.C. did the analyses. F.C.C. prepared the figures. All authors reviewed the manuscript.
{#f1}
{#f2}
{#f3}
###### Median attack ratio and 95% credibility intervals calculated according to (13).
^†^Year ^‡^Cases median Attack Ratio ^‡^*S*~0~ Circulating Serotypes
--------- ---------- ---------------------- -------------------- -----------------------
1996 0.066 0.39 (0.17--0.54) 0.171 (0.12--0.38) DEN-1 and 2[@b13]
1997 0.238 0.87 (0.74--0.87) 0.273 (0.27--0.32) DEN-1 and 2[@b13]
1998 0.0708 0.5 (0.49--0.5) 0.142 (0.14--0.14) DEN-1 and 2[@b13]
1999 0.0371 0.11 (0.037--0.2) 0.345 (0.18--1.0) DEN-1, 2 and 3[@b14]
2000 0.394 0.25 (0.24--0.27) 1.55 (1.5--1.6) DEN-1, 2 and 3[@b14]
2001 2.38 0.48 (0.47--0.49) 4.95 (4.8--5.1) DEN-1, 2 and 3[@b14]
2005 0.217 0.15 (0.1--0.21) 1.47 (1.0--2.1)
2006 0.315 0.11 (0.08--0.14) 2.81 (2.2--3.7) DEN-2[@b15]
2007 2.03 0.15 (0.15--0.15) 13.5 (13.0--14.01) DEN-2[@b15]
2008 0.0923 0.14 (0.031--0.31) 0.672 (0.3--2.4) DEN-2[@b15]
2010 0.831 0.18 (0.17--0.19) 4.54 (4.3--4.8) DEN-2[@b15]
2011 1.85 0.086 (0.082--0.094) 21.5 (20.0--23.0) DEN-1, 2 and 4[@b15]
2012 0.864 0.14 (0.13--0.15) 6.21 (5.8--6.8)
Values are presented as percentage of total population. ^†^Year corresponds to the start of the epidemic, but the peak of cases may occur in the following year. ^‡^percentage of population. These results show considerable variation in AR between epidemics, consistent with the accquiring and loss of serotype-specific immunity.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
The coastal marshes of Louisiana are experiencing sea-level rise at rates of approximately 10--11 mm/year; these rates are 10 times the global averages ([@b15]; [@b9]). During the last 50 years, Louisiana has lost coastal lands at a rate of over 130 km^2^ per year. This results in stands of marsh vegetation separated by open water ([@b18]; [@b16]). As marsh areas are converted to shallow open water areas, saline waters can penetrate farther up the estuary because there is less resistance from the marsh to up-estuary flows from tidal flux and storm surge.
Fish inhabiting these deteriorating coastal marshes may be especially susceptible to salinity stress when hurricanes and other storms push salt water into lower salinity marshes. Three nonexclusive mechanisms may govern the ability of fish to survive salinity spikes, or more permanent salinity increases caused by sea-level rise and canalization, and thus maintain viable populations. First, coastal marsh residents may be euryhaline, having an ability to tolerate a wide salinity range. Second, acclimatization to brackish conditions may make fish less susceptible to salinity increases than fish acclimatized to fresh water. Finally, fish living in brackish conditions may acquire genetic adaptations through natural selection for higher individual salinity tolerance than fish from freshwater environments where there has been no selection for salinity tolerance.
Based on previous sampling, the most abundant fish species in coastal freshwater marshes of Louisiana is the Western mosquitofish *Gambusia affinis* (Baird & Girard, 1853). This marsh resident also occurs in brackish marshes, although in reduced numbers ([@b13]). Because of the broad range of salinity across which this species occurs, we chose it as a model to evaluate the impact of saltwater intrusion on coastal marsh fish populations. *Gambusia affini*s is a small, sexually dimorphic Poeciliid with a maximum size of 40--50 mm standard length; females are generally larger than males ([@b6]).
Salinity tolerance of *G. affinis* and its congener *Gambusia holbrooki* have been examined extensively in previous studies ([@b8]; [@b4]; Kandl and Thompson 1996; [@b14]). These studies clearly show that *Gambusia* has a broad physiological tolerance for salinity and that laboratory acclimation to increased salinity is possible*.* There is also evidence that frequencies of allozyme genotypes differ between fresh and more saline conditions ([@b5]; [@b11]), suggesting that genetic adaptation to different levels of salinity is possible. In spite of this large body of research, it is not clear whether field populations are adapted to increased salinity levels in coastal marshes, or if fish survival is more dependent on having a wide physiological tolerance. The latter would be an advantage in the dynamic environment of coastal marshes where salinities can change rapidly and unpredictably. Rapid changes in salinities would make acclimatization difficult and would be even less favorable to adaptation that might require many generations. The possibility that individuals may move across salinity gradients, resulting in gene flow, would further work against adaptation to localized salinities ([@b26]).
Our objective was to determine if populations of *G. affinis* with a history of exposure to brackish water would display a different level of survival relative to populations without a history of salinity exposure. In addition, we examined if the saltwater intrusion in coastal marshes could serve as a mechanism driving selection for genetic adaptations to increased salinity tolerance in a common marsh resident. The mechanisms through which fish tolerate increases in salinity have implications for the ability of populations to survive short-term changes in salinity due to storm surges, as well as longer term changes due to relative sea-level rise.
Methods and materials
=====================
Salinity varies temporally in coastal marshes; so, salinity measured at any point in time may not reflect long-term conditions at a site. We selected sample sites based on vegetation classification ([@b29]) because the different marsh types in the classification system reflect average salinity conditions that occurred in the recent past. Three marsh types were chosen for fish collections: fresh (0--1‰), intermediate (3--5‰) and brackish (8--12‰). Collections were made along a salinity gradient in Cameron Parish in south-western Louisiana and another gradient in Terrebonne Parish in south-eastern Louisiana. These collection sites were located approximately 333 km apart in an attempt to see if the relationship between salinity tolerance and past exposure history were similar across space, or a function of local conditions.
Collected fish were returned to the laboratory where source populations were placed in separate 2800-L freshwater tanks ([@b25]). All fish were acclimated to freshwater for a period of 1 week to ensure any effect due to past salinity exposure was not due to salinity experienced immediately prior to the experimental exposure ([@b20]).
From the acclimation tanks, we selected 100 individuals from each marsh type in both south-western and south-eastern Louisiana (total *n* = 600). For the exposure experiment, each of the 600 fish was placed in a separate 450-mL chamber. These chambers were large enough to prevent oxygen depletion or ammonia toxicity for at least 8 h (K. M. Purcell, unpublished data). A 50% water change was performed every 8 h during the study to prevent stress from anoxia or ammonia toxicity. All chambers were covered with screen and maintained at a temperature of 22.7°C.
The water in chambers of 70 individuals from each of the source populations was spiked with 40 ml of a concentrated artificial salt solution to increase the water salinity to 25‰. The other 30 individuals from each source population were held in control chambers and were spiked with equal volumes of fresh water. The exposure began following inoculation of all experimental chambers. Fatalities were counted at 30 min intervals and death was defined as a lack of opercular movement and no response to three contacts with a blunt instrument ([@b7]). Dead individuals were removed from their chamber; time of death, sex and standard length were recorded.
To determine if there was a genetic basis for differences in salinity resistance based on past exposure, we evaluated tolerance using descendents of fish from fresh and brackish marshes from each location. Fish from the same fresh and brackish marshes sampled in the earlier trials were maintained in freshwater conditions and allowed to reproduce. Their descendents were reared in freshwater for a minimum two generations. Therefore, we were able to assess salinity resistance for individuals that had never experienced saline conditions and whose parents had never experienced such conditions. By using fish reared for two generations in a 'common garden', we prevented the effect of salinity acclimatization and the effect of maternal exposure to salinity ([@b19; @b20]), from influencing salinity resistance. Any differences in resistance associated with historical salinity exposure would be best explained through adaptation of field populations to local conditions. The procedures of these salinity exposure trials were the same as those described earlier using fish sampled from the coastal marshes.
All statistical analyses were conducted using SAS software ([@b17]). The null hypotheses were that there is no difference in the time to death of individuals collected from different salinities, or their descendents, when exposed to salinity stress. A proportional hazards model was chosen for these analyses because it is a robust, semi-parametric method that does not require data to fit a specific probability distribution ([@b2]). We assessed the effects of sex and standard length of individuals on survival time as covariates in all trials. These covariates may be collinear, because females obtain greater lengths than males. Therefore, the effect of each covariate was assessed with and without the inclusion of the other covariate in the analysis. The PHREG procedure ([@b2]) was used to determine the effects of source population, sex and length on survival time. The LIFETEST procedure ([@b2]) was used to estimate the survival distribution function for each of the source populations.
Results
=======
In the comparison of fish from Cameron Parish exposed to a salinity of 25‰, the survival time of individuals from freshwater marshes was significantly less than fish from marshes with a history of salinity exposure (*χ*~1~^2^ = 26.16, *P* \< 0.0001; [Fig. 1](#fig01){ref-type="fig"}). This difference appeared to be due to reduced survival times of fish from freshwater marshes; the survival times of fish from brackish and intermediate marshes were similar. The survival time increased proportionally with increases in the standard length of individuals (*χ*~1~^2^ = 7.38, *P* = 0.0066; [Fig. 2](#fig02){ref-type="fig"}). For the Cameron Parish fish, sex did not have a significant effect on survival, regardless of whether standard length was included in the analysis.
{#fig01}
{#fig02}
For fish from Terrebonne Parish, survival time differed significantly among the three populations with different historical salinity exposure (*χ*~1~^2^ = 40.85, *P* \< 0.0001; [Fig. 1](#fig01){ref-type="fig"}). Individuals from freshwater marsh had the least salinity tolerance; those from saline marshes were the most tolerant ([Fig. 1](#fig01){ref-type="fig"}). In this comparison, the standard length of individuals was not found to have a significant effect on the survival time, regardless of whether sex was included in the analysis. Sex, however, was found to have a significant effect on survival time (*χ*~1~^2^ = 12.91, *P* = 0.0003); females had a higher mean survival time than males ([Fig. 2](#fig02){ref-type="fig"}).
The descendents of fish from fresh and brackish marshes in Cameron Parish had significantly different survival times (*χ*~1~^2^ = 14.65, *P* \< 0.0001; [Fig. 3](#fig03){ref-type="fig"}). The descendents of individuals from freshwater marsh again displayed reduced survival time relative to fish from brackish marshes when exposed to 25‰ ([Fig. 3](#fig03){ref-type="fig"}). As in the parental generation survival time was significantly affected by standard length (*χ*~1~^2^ = 11.04, *P* \< 0.0009). Increases in the standard length of individuals were positively related to increased survival under stress. The covariate sex did not have a significant impact on survival time in this trial regardless of the inclusion of other covariates.
{#fig03}
When comparing the fish from the Terrebonne Parish sites, we found that there was a significant difference between individuals descended from the fresh and brackish populations when exposed to salinity stress (*χ*~1~^2^ = 29.65, *P* \< 0.0001; [Fig. 3](#fig03){ref-type="fig"}). The descendents of fish from the freshwater marsh had a considerably reduced survival time relative to those descended from fish from brackish marsh ([Fig. 3](#fig03){ref-type="fig"}). In this trial, neither of the covariates had any significant effects on the survival time of individuals.
Discussion
==========
Individuals from populations of *G. affinis* with historical exposure to salinity experienced increased survival during elevated salinities. Fish collected from brackish marsh habitats had significantly greater survival than those collected from habitats without historical salinity exposure when treated with a salinity stress of 25‰. Clearly, the wide range of salinity tolerance known to occur in marsh residents like *Gambusia* ([@b1]; [@b4]; [@b13]) is influenced strongly by historical salinity exposure.
There would appear to be only three possible explanations for this outcome in field-collected fish. First, individuals were acclimatized to the salinity of their capture site. Such acclimatization is typically thought to be reversible; so, exposing the fish to freshwater for a period of 1 week prior to experimental exposure should have minimized the effect of acclimatization. In another poecilliid, *Poecilia reticulata*, saltwater tolerance was the same for fish raised at high salinities (acclimatization) as for fish acclimated to high salinities for only a short time ([@b19]). Thus, if acclimatization is responsible for our result, it is less reversible in *Gambusia* than in other fish.
Second, it is possible that fish from the different marsh types varied in their sex or size composition that would result in differences in salinity tolerances. We attempted to control for this possibility in field-collected fish by including two covariates, sex and standard length, in our analyses. The sexual dimorphism of the model species *G. affinis* would suggest that any relationship of these two variables with salinity tolerance would be collinear, with females being larger than males ([@b6]). We hypothesized that as an individual fish increases in size, its tolerance for saline environments also increases due mainly to its increased mass; similar relationships have been seen in other studies of *Gambusia* tolerance to stress ([@b4]; [@b7]; [@b14]). The covariates had a different effect on survival time depending on the location of the collection. This result indicated that neither sex nor standard length could be used as a reliable estimator of survival following a salinity stress event. The only consistent factor across locations was that past exposure to salinity increased tolerance; the effects of exposure could not be explained by variation in size or sex structure of fish collected from the field sites.
A third possible explanation is that individuals collected from populations with a history of salinity exposure have adaptations that promote survival in high-salinity environments. Genetic adaptation to salinity tolerance has been documented in laboratory strains of *P. reticulata* ([@b21]; [@b22; @b23; @b24]). Here, we use the term 'adaptation' to refer to a tolerance advantage with a genetic basis resulting from a population being exposed to elevated environmental salinities for several generations. The evidence for adaptations in salinity is not clear for natural populations of *Gambusia.* Studies of *G. holbrooki* under salinity stress showed differential survival among allozyme genotypes ([@b5]) and genotypic variation might be related to performance in mesocosms under salinity stress ([@b11]); however, these studies did not demonstrate local adaptation to salinity levels. [@b26] conclude that a population of *G. affinis* from a small freshwater stream might have been maladapted for those conditions because it apparently was heavily influenced by gene flow from fish from a large population in a brackish marsh. They saw no evidence that selection for local salinity levels had resulted in local adaptation and we are aware of no investigations of adaptation to localized salinity levels in *Gambusia*.
Our results indicate that individuals that are the descendents of populations with historical salinity exposure, reared in fresh water, are more resistant to salinity stress than the offspring of populations without historical exposure. This difference in salinity tolerance among individuals that had never experienced elevated salinity strongly suggest a genetic basis for the differences observed between fish from brackish and freshwater marshes. This genetic basis for increases in salinity tolerance to stress events suggests that at least some resident marsh fishes have the capacity to respond to saltwater intrusion through natural selection.
Assuming the presence of genetic variation, the most important factor-limiting development of local adaptation to salinity in coastal environments would be gene flow. Gene flow plays an important role in the contemporary evolution of populations ([@b27]); however, it also has the potential to prevent adaptation to localized conditions along gradients such as those existing in coastal marshes. [@b12] reviews of both theoretical and empirical studies demonstrating how gene flow into populations alters gene frequencies and can limit a population's ability for adaptation. Closer to our study, [@b26] documented an example in which gene flow from a large population of *G. affinis* prevented a much smaller population from adapting to local salinity conditions. The populations in our study were separated by greater distances than those reported in Stearns and Sage, and were in some cases separated by physical barriers such as levees, weirs and open water with no aquatic vegetation. These barriers may have restricted gene flow sufficiently to allow for local adaptation in the *G. affinis* populations we examined. It is an open question as to whether gene flow is sufficiently high in other marsh residents to prevent local adaptation to increasing salinity.
The adaptation of populations in contemporary time frames to local conditions is of growing interest to ecologists ([@b27]; [@b3]; [@b28]). It is interesting that we see evidence for a genetic basis for adaptations along two salinity gradients separated by 330 km, a distance that would require genes many generations to transverse. Furthermore, these sample sites are isolated by a major barrier, the Atchafayala River. This large river would make direct gene flow between the two populations experiencing brackish conditions impossible; populations along the river would most likely be adapted to freshwater conditions. Although this freshwater barrier and great distance suggests that salinity tolerance may have evolved due to the local conditions of the sample sites, available evidence provides no insight into whether the adaptations to increased salinity are the same in these two coastal marshes. The degree to which populations can adapt to changing salinities, will dictate how coastal fish communities respond to both short-term phenomena like storm surges from hurricanes, as well as long-term changes due to relative sea-level rise.
We thank EPA National Center of Environmental Research and the Louisiana Board of Regents EPSCor program for funding this project. The experiments would not have been possible without the assistance of K. Barr, N. Giridhar, C. Green, S. Martin, L. McDonald, J. McGettigan and E. Smithies. The manuscript was improved by the comments of R. Bonett, J. Neigel, L. Rozas and R. Twilley. This research complies with all legal requirements and was conducted under IACUC \# 2004-8717-031.
[^1]: **Present address:** School of Forestry and Wildlife Sciences, Auburn University, 602 Duncan Dr, Auburn, AL 36849-5814, USA.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
Cardiac disease is a major cause of mortality in modern society. The high prevalence of obesity and related diseases, such as lipotoxic cardiomyopathy, plays a significant role in the increased incidence of heart failure. Heart failure affects more than a billion people worldwide. Lipotoxic cardiomyopathy is caused by excessive lipid accumulation in myocardial cells, and it is a form of cardiac dysfunction ([@BIO044719C9]). For example, high-fat diet (HFD)-fed flies exhibit increased triglyceride (TG) fat and alterations in insulin/glucose homeostasis, similar to mammalian responses. A HFD also causes cardiac lipid accumulation, cardiac contractility reduction, conduction blocks and severe structural pathologies, reminiscent of diabetic cardiomyopathies ([@BIO044719C8]). In addition, increasing evidence shows that inhibition of signaling through insulin, target of rapamycin (TOR), or the lipogenic transcription factor SREBP, or by increasing TG lipolysis, is effective in counteracting excess lipid accumulation as well as the associated cardiac defects in flies ([@BIO044719C8]; [@BIO044719C17]; [@BIO044719C27]). These reports suggest that both a HFD and cardiac lipid metabolism genes are closely related to lipotoxic cardiomyopathy.
*Sir2* is the most intensively discussed longevity gene in current aging research. Importantly, some studies have shown that *Sir2* is involved in lipid metabolism regulation; a screen for obesity-inducing genes in *Drosophila* larvae pointed to a role for *Sir2* in regulating fat metabolism and a response to amino-acid starvation ([@BIO044719C53]). Moreover, *Sir2* apparently regulates expression of genes involved in fat metabolism, and the lack of *Sir2* increases fat deposition under normal conditions and consequently impairs starvation survival of flies ([@BIO044719C7]). Next, a recent study shows HFD-fed flies exhibit increased body TG levels and decreased body *dSir2* expression ([@BIO044719C70]). However, it is unclear whether *Sir2* can take part into heart lipid metabolism regulation. A recent study has confirmed *PGC-1α* as a vital antagonist of HFD-induced lipotoxic cardiomyopathy in flies since it plays key roles in mitochondrial biogenesis and electron transport chain assembly ([@BIO044719C19]; [@BIO044719C20]). Interestingly, *Sir2* expression can alter the transcriptional activity of the mitochondrial biogenesis coactivator PGC-1α, and it catalyzes PGC-1α deacetylation both *in vitro* and *in vivo*. Overexpression of Sir2 deacetylase or increasing NAD^+^ levels activates transcriptional activity of PGC-1α in neurons and increases mitochondrial density ([@BIO044719C16]; [@BIO044719C41]). Therefore, studying the relationship between cardiac NAD^+^/dSIR2/*PGC-1α* pathway and heart lipid metabolism is very important to understand the mechanism of lipotoxic cardiomyopathy formation.
In modern society, exercise combined with a healthy diet is considered the most economical and non-invasive way to prevent and treat obesity ([@BIO044719C6]). Exercise also improves heart function and decreases incidence of heart failure in both human and *Drosophila*. Recent studies report that endurance exercise improves cardiac contraction, and it reduces body and heart lipid levels and heart fibrillation in both HFD and aging flies ([@BIO044719C70]; [@BIO044719C77]). Similarly, exercise training alters extrinsic modulation of the heart and improves the intrinsic pump capacity of the heart in human, and it also improves quality of life of patients with chronic heart failure ([@BIO044719C15]; [@BIO044719C32]; [@BIO044719C68]). In addition, exercise increased muscle NAD^+^ levels ([@BIO044719C25]; [@BIO044719C38]) and increasing NAD^+^ levels activates transcriptional activity of PGC-1α in neurons and increases mitochondrial density ([@BIO044719C16]; [@BIO044719C41]). Therefore, these results suggest that NAD^+^/dSIR2/*PGC-1α* pathway activation may be one of the key mechanisms in which exercise improves heart function and prevents lipotoxic cardiomyopathy.
In this study, to explore whether endurance exercise could resist HFD-induced lipotoxic cardiomyopathy via activating NAD^+^/dSIR2/*PGC-1α* signal pathway, experimental flies were fed a HFD and given exercise, and heart *dSir2* expression was changed by building UAS/hand-Gal4 system. The heart TG levels and *dFAS* gene expression were reflected in the lipid metabolism status. In addition, the heart diastolic diameter, systolic diameter, fractional shortening and arrhythmia index were reflected in the heart function. Finally, the cardiac NAD^+^ levels, dSIR2 protein level, *dSir2* mRNA expression and *PGC-1α* mRNA expression were reflected in the heart NAD^+^/dSIR2/ *PGC-1α* pathway status.
RESULTS {#s2}
=======
Exercise prevented lipotoxic cardiomyopathy and activated cardiac NAD^+^/dSIR2/*PGC-1α* pathway in *Drosophila* {#s2a}
---------------------------------------------------------------------------------------------------------------
Increasing evidence confirms that heart lipotoxicity impairment can be induced by feeding HFD in both mammals and flies, and the heart contractility and ejection fraction were weakened at the same time ([@BIO044719C1]; [@BIO044719C59]). Lipotoxicity impairment is also accompanied by the dysfunction of some genes in the heart, such as *PGC-1α*, *dFAS*, *TOR*, etc. ([@BIO044719C8]; [@BIO044719C17]; [@BIO044719C19]). On the contrary, exercise training improves heart function, decreases incidence of heart failure in both mammals and *Drosophila*, and reduces body and heart fat levels and heart fibrillation ([@BIO044719C70]; [@BIO044719C77]). In addition, exercise increased muscle NAD^+^ levels ([@BIO044719C25]; [@BIO044719C38]), and the increasing NAD^+^ levels could activate transcriptional activity of PGC-1α in cells and increase mitochondrial density ([@BIO044719C16]; [@BIO044719C41]). These results hinted that NAD^+^/dSIR2/*PGC-1α* pathway activation may be one of key mechanisms that exercise improved heart function and prevented lipotoxic cardiomyopathy. To identify this hypothesis, fruit flies in this experiment were subjected to exercise intervention and a HFD intervention.
In this study, our results showed that a HFD remarkably increased heart TG levels in untrained-*w^1118^* flies (*P*\<0.01), and it also upregulated heart *dFAS* expression levels (*P*\<0.01). These were consistent with the results of previous studies. In addition, we found that exercise availably reduced heart TG level and *dFAS* expression level in both *w^1118^*-normal diet (ND) and *w^1118^*-HFD flies (*P*\<0.01, *P*\<0.05). Interestingly, the heart TG levels in *w^1118^*-high-fat diet+exercise (HFD+E) flies were lower than those in *w^1118^*-ND flies (*P*\<0.05). Therefore, we identified that endurance exercise could prevent lipid accumulation by downregulating cardiac *dFAS* expression level ([Fig. 1](#BIO044719F1){ref-type="fig"}A,B). Fig. 1.**The effect of a HFD and exercise on heart lipid accumulation.** (A) Heart relative TG level. Results are expressed as the fold difference compared with *w^1118^*-ND flies. The sample size was 80 hearts with three biological replicates. (B) Heart *dFAS* expression level. The sample size was 80 hearts with three biological replicates. A two-way ANOVA was used to identify differences among the ND, ND+E, HFD, and HFD+E groups in *w^1118^* flies. Data are represented as means±s.e.m. \**P*\<0.05; \*\**P*\<0.01.
For heart function, results displayed that a HFD significantly reduced heart fractional shortening (FS) in untrained *w^1118^* flies (*P*\<0.01), and it also notably decreased heart diastolic diameters in untrained *w^1118^* flies (*P*\<0.05). Exercise significantly increased FS in both *w^1118^*-HFD flies and *w^1118^*-ND flies (both *P*\<0.05), and it also increased heart diastolic diameters in both *w^1118^*-HFD flies and *w^1118^*-ND flies (both *P*\<0.05). Importantly, there was no significant difference between *w^1118^*-HFD+E flies and *w^1118^*-ND flies in FS (*P*\>0.05) ([Fig. 2](#BIO044719F2){ref-type="fig"}A--C). Moreover, a HFD significantly increased arrhythmia index (AI) in untrained *w^1118^* flies (*P*\<0.05). Exercise reduced AI in *w^1118^*-HFD flies (*P*\<0.05). There was no significant difference between *w^1118^*-HFD+E flies and *w^1118^*-ND flies in AI (*P*\>0.05) ([Fig. 2](#BIO044719F2){ref-type="fig"}D). These results confirmed that a HFD could weaken heart contractility and increase the risk of arrhythmia in *w^1118^* flies, but endurance exercise could prevent this from happening in a HFD heart ([Fig. 2](#BIO044719F2){ref-type="fig"}E1--E4). Fig. 2.**The effect of a HFD and exercise on heart function.** (A) Heart diastolic diameters. (B) Heart systolic diameters. (C) Fractional shortening. (D) Arrhythmia index. (E) Illustrating qualitative differences in heart function parameters (10 s): fractional shortening and arrhythmia index; E1: *w^1118^*-ND; E2: *w^1118^*-ND+E; E3: *w^1118^*-HFD; E4: *w^1118^*-HFD+E. A two-way ANOVA was used to identify differences among the ND, ND+E, HFD, and HFD+E groups in *w^1118^* flies. Data are represented as means±s.e.m. \**P*\<0.05; \*\**P*\<0.01. The sample size was 30 hearts.
The results also showed that a HFD significantly reduced cardiac NAD^+^ level, dSIR2 level, heart *dSir2* expression and *PGC-1α* expression level in untrained *w^1118^* flies (*P*\<0.05, *P*\<0.01). Exercise significantly increased cardiac NAD^+^ level, dSIR2 level, heart *dSir2* expression and *PGC-1α* expression level in both *w^1118^*-HFD flies and *w^1118^*-ND flies (*P*\<0.01). Importantly, the cardiac *PGC-1α* expression levels in *w^1118^*-HFD+E flies was higher than that of *w^1118^*-ND flies (*P*\<0.05) ([Fig. 3](#BIO044719F3){ref-type="fig"}A--D). Since the *PGC-1α* was involved in the synthesis of mitochondria, the number and morphology of mitochondria in cardiac cells was determined by transmission electron microscopy. We observed that in both HFD flies and non-HFD flies, exercise increased mitochondrial numbers and improved myofibril arrangement regularity in myocardial cells ([Fig. 3](#BIO044719F3){ref-type="fig"}E1--E4). Therefore, these results confirmed that a HFD induced a decreased in heart NAD^+^/dSIR2/*PGC-1α* pathway activity, but exercise training could prevent this from happening and even improve heart NAD^+^/dSIR2/*PGC-1α* pathway in HFD flies. Fig. 3.**The effect of a HFD and exercise on NAD^+^/dSIR2/*PGC-1α* pathway.** (A) Heart NAD^+^ levels. Results are expressed as the fold difference compared with *w^1118^*-ND flies. (B) Heart dSIR2 level. Results are expressed as the fold difference compared with *w^1118^*-ND flies. (C) Heart *dSir2* mRNA expression level. (D) Heart *PGC-1α* mRNA expression level. (E) The images of transmission electron microscopy; E1: *w^1118^*-ND; E2: *w^1118^*-ND+E; E3: *w^1118^*-HFD; E4: *w^1118^*-HFD+E. As we have seen exercise increased the mitochondrial numbers and improved myofibril arrangement regularity in myocardial cells. The red circle represented the position of the mitochondria in myocardial cells. The arrow represented the position of the Z line in myocardial cells. A two-way ANOVA was used to identify differences among the ND, ND+E, HFD, and HFD+E groups in *w^1118^* flies. Data are represented as means±s.e.m. \**P*\<0.05; \*\**P*\<0.01. The sample size of these indicators was 80 hearts, with three biological replicates.
Lipotoxic cardiomyopathy and cardiac *dSir2* gene in *Drosophila* {#s2b}
-----------------------------------------------------------------
Sir2 is an archetypal longevity gene and founder of the Sirtuin protein family, and it is involved in the regulation of rDNA stability and the control of cellular lifespan and aging ([@BIO044719C11]; [@BIO044719C35]; [@BIO044719C60]; [@BIO044719C72]). Recent studies report that *dSir2* also takes part into the regulation of lipid metabolism ([@BIO044719C28]; [@BIO044719C70]). For instance, overexpression of *dSir2* restricts fat accumulation in individual cells of the fat body in a cell autonomous manner, and loss of the *dSir2* leads to the age-progressive onset of hyperglycemia, obesity, glucose intolerance and insulin resistance ([@BIO044719C50]; [@BIO044719C53]). However, it remains unclear how *dSir2* regulates lipid metabolism. Some evidence suggests that *Sir2* expression can alter the transcriptional activity of the mitochondrial biogenesis coactivator PGC-1α, and it catalyzes PGC-1α deacetylation both *in vitro* and *in vivo*. Overexpression of Sir2 deacetylase or increasing NAD^+^ levels activate transcriptional activity of *PGC-1α* in neurons and increases mitochondrial density ([@BIO044719C16]; [@BIO044719C41]). It has been reported the *PGC-1α* in fly heart is a key gene in regulating the formation of lipotoxic cardiomyopathy ([@BIO044719C19]; [@BIO044719C20]). To confirm the guess that heart *dSir2* gene can regulate cardiac lipid metabolism via modulating NAD^+^/dSIR2/*PGC-1α* pathway, the cardiac *dSir2* gene was overexpressed and knocked down by building UAS/hand-Gal4 system in *Drosophila*.
Cardiac *dSir2* overexpression reduced the risk of lipotoxic cardiomyopathy {#s2c}
---------------------------------------------------------------------------
In this research, the results showed the cardiac *dSir2* mRNA expression of *dSir2*-overexpression-normal diet (*dSir2*-OE-ND) flies was higher than that of *dSir2*-control flies (*P*\<0.01, about 3.1-fold higher) ([Fig. 4](#BIO044719F4){ref-type="fig"}A). This suggested that cardiac *dSir2* gene overexpression was successfully constructed. Since a dnaJ-homologue (*dnaJ-H*) gene partially overlaps with *dSir2*, the cardiac *dSir2* overexpression might affect cardiac *dnaJ-H* mRNA expression. To avoid the influence of cardiac *dnaJ-H* mRNA expression on our results, we measured the expression of *dnaJ-H* gene in the heart*.* The results showed that there was no significant difference between *dSir2*-control flies and *dSir2*-OE flies in cardiac *dnaJ-H* mRNA expression (*P*\>0.05) ([Fig. 4](#BIO044719F4){ref-type="fig"}K). Increasing evidence indicates that moderately increased expression of *dSir2* (2.5-fold, threefold, and fivefold) from the native *dSir2* locus cannot increase *dnaJ-H* mRNA expression ([@BIO044719C11]; [@BIO044719C28]). In this study, cardiac *dSir2* mRNA expression of *dSir2*-OE-ND flies was about 3.1-fold higher than that of *dSir2*-control flies. This hinted that cardiac *dSir2* overexpression did not significantly affect cardiac *dnaJ-H* mRNA expression, and cardiac *dnaJ-H* gene may not affect our results in this study. Also, results showed that cardiac *dSir2* overexpression significantly increased heart dSIR2 level, NAD^+^ level, and *PGC-1α* expression level (*P*\<0.05, *P*\<0.05, *P*\<0.01) when *dSir2*-OE-ND flies were compared to *dSir2*-control flies ([Fig. 4](#BIO044719F4){ref-type="fig"}B--D). It hinted that cardiac *dSir2* gene overexpression upregulated NAD^+^/dSIR2/*PGC-1α* pathway activity. Moreover, we found heart diastolic diameter and fractional shortening of *dSir2*-OE-ND flies were higher than that of *dSir2*-control flies (*P*\<0.05) ([Fig. 4](#BIO044719F4){ref-type="fig"}E,G). The arrhythmia index of *dSir2*-OE-ND flies was lower than that of *dSir2*-control flies (*P*\<0.05) ([Fig. 4](#BIO044719F4){ref-type="fig"}H,L). This suggested that cardiac *dSir2* gene overexpression could enhance cardiac contractility and decrease the risk of arrhythmia. Finally, we found that the heart TG level and *dFAS* expression of *dSir2*-OE-ND flies was lower than that of *dSir2*-control flies (*P*\<0.01) ([Fig. 4](#BIO044719F4){ref-type="fig"}-I,J). Therefore, these results indicated that cardiac *dSir2* gene overexpression could prevent lipid accumulation in the heart. According to others and our results, we hypothesized that cardiac *dSir2* gene overexpression could reduce the risk of lipotoxic cardiomyopathy via activating cardiac NAD^+^/dSIR2/ *PGC-1α* pathway in old flies. Fig. 4.**The effect of a HFD and exercise on lipotoxic cardiomyopathy in *dSir2*-OE flies.** (A) Relative heart *dSir2* mRNA expression level. (B) Heart dSIR2 level. (C) Heart NAD^+^ levels. (D) Relative heart *PGC-1α* mRNA expression level. (E) Heart diastolic diameter. (F) Heart systolic diameter. (G) Fractional shortening. (H) Arrhythmia index. (I) Heart TG levels. Results are expressed as the fold difference compared with *dSir2-*control flies. (J) Relative heart *dFAS* gene expression level. (K) Relative heart *dnaJ-H* mRNA expression level. (L) Illustrating qualitative differences in heart function parameters (6 s): fractional shortening and arrhythmia index; L1: *dSir2*-control; L2: *dSir2*-OE-ND; L3: *dSir2*-OE-ND+E; L4: *dSir2*-OE-HFD; L5: *dSir2*-OE-HFD+E. Independent sample *t*-test was used to identify differences between *dSir2*-control flies and *dSir2*-OE flies. A two-way ANOVA was used to identify differences among the ND, ND+E, HFD, and HFD+E groups in *dSir2*-OE flies. Data are represented as means±s.e.m. \**P*\<0.05; \*\**P*\<0.01. The sample size was the same as with *w^1118^*.
To further confirm whether cardiac *dSir2* gene overexpression could prevent lipotoxic cardiomyopathy induced by a HFD, the *dSir2*-OE flies were fed a HFD. Results showed that the cardiac *dSir2* expression level, dSIR2 level, NAD^+^ level, *PGC-1α* expression level, diastolic diameter, fractional shortening, arrhythmia index, heart TG level and *dFAS* expression of *dSir2*-OE-ND flies were not significantly different from that of *dSir2*-OE-HFD flies (*P*\>0.05) ([Fig. 4](#BIO044719F4){ref-type="fig"}). These results confirmed that cardiac *dSir2* gene overexpression could prevent lipotoxic cardiomyopathy induced by a HFD in old flies. The mechanism is that a HFD cannot weaken the activity of NAD^+^/dSIR2/*PGC-1α* pathway in *dSir2*-OE flies.
In *w^1118^* flies, we have confirmed that endurance exercise could resist lipotoxic cardiomyopathy and activate NAD^+^/dSIR2/*PGC-1α* pathway, but the relationship between endurance exercise and cardiac *dSir2* gene overexpression in preventing lipotoxic cardiomyopathy induced by a HFD remained unknown. So, the *dSir2*-OE flies participated in exercise training. We found that endurance exercise significantly upregulated the expression of cardiac *dSir2* gene in both *dSir2*-OE-ND flies and *dSir2*-OE-HFD flies (*P*\<0.01 and *P*\<0.05, respectively) ([Fig. 4](#BIO044719F4){ref-type="fig"}A), and it also remarkably increased heart dSIR2 level, NAD^+^ level and *PGC-1α* expression level in both *dSir2*-OE-ND flies and *dSir2*-OE-HFD flies (*P*\<0.05 and *P*\<0.01, respectively) ([Fig. 4](#BIO044719F4){ref-type="fig"}B--D). In addition, endurance exercise significantly increased diastolic diameter in *dSir2*-OE-ND flies (*P*\<0.05), and it significantly increased fractional shortening in both *dSir2*-OE-ND flies and *dSir2*-OE-HFD flies (*P*\<0.055) ([Fig. 4](#BIO044719F4){ref-type="fig"}E--G). Moreover, endurance exercise significantly reduced heart TG level and *dFAS* expression in both *dSir2*-OE-ND flies and *dSir2*-OE-HFD flies (*P*\<0.05 and *P*\<0.01, respectively) ([Fig. 4](#BIO044719F4){ref-type="fig"}I,J). Therefore, in the fight against lipotoxic cardiomyopathy, overexpression of cardiac *dSir2* was parallel to exercise training. Overexpression of cardiac *dSir2* combined with exercise training could better prevent lipotoxic cardiomyopathy in old flies. The mechanism was that both overexpression of cardiac *dSir2* and exercise training could superimpose the improvement of the activity of cardiac NAD^+^/dSIR2/*PGC-1α* pathway.
Exercise improved lipotoxic cardiomyopathy induced by cardiac *dSir2* knockdown {#s2d}
-------------------------------------------------------------------------------
To further confirm the relationship between cardiac *dSir2* and lipotoxic cardiomyopathy, it was necessary to construct cardiac *dSir2* knockdown (KD) by UAS/hand-Gal4 system. In this study, the results showed that the cardiac *dSir2* mRNA expression of *dSir2*-KD-ND flies was lower than that of *dSir2*-control flies (*P*\<0.01, about 2.5-fold lower) ([Fig. 5](#BIO044719F5){ref-type="fig"}A). It suggested that cardiac *dSir2* gene knockdown was successfully constructed. In addition, when *dSir2*-OE-ND flies were compared to *dSir2*-control flies, we found cardiac *dSir2* knockdown significantly decreased heart dSIR2 levels, NAD^+^ levels, and *PGC-1α* expression levels (*P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}B--D). This suggested that cardiac *dSir2* gene knockdown inhibited NAD^+^/dSIR2/*PGC-1α* pathway activity. Moreover, we found the heart diastolic diameter and fractional shortening of *dSir2*-KD-ND flies were lower than that of *dSir2*-control flies (*P*\<0.05, *P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}E,G). The arrhythmia index of *dSir2*-KD-ND flies was higher than that of *dSir2*-control flies (*P*\<0.05) ([Fig. 5](#BIO044719F5){ref-type="fig"}H,K). This suggested that cardiac *dSir2* gene knockdown could weaken cardiac contractility and increase the risk of arrhythmia. Finally, we found that the heart TG levels and *dFAS* expression of *dSir2*-KD-ND flies were higher than that of *dSir2*-control flies (*P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}I,J). This indicated that cardiac *dSir2* gene knockdown could induce lipid accumulation in the heart. Therefore, according to others' and our own evidence, we hypothesized that cardiac *dSir2* gene knockdown could induce lipotoxic cardiomyopathy via inhibiting cardiac NAD^+^/dSIR2/ *PGC-1α* pathway in old flies. Fig. 5.**The effect of a HFD and exercise on lipotoxic cardiomyopathy in *dSir2*-KD flies.** (A) Relative heart *dSir2* mRNA expression level. (B) Heart dSIR2 level. (C) Heart NAD^+^ levels. (D) Relative heart *PGC-1α* mRNA expression level. (E) Heart diastolic diameter. (F) Heart systolic diameter. (G) Fractional shortening. (H) Arrhythmia index. (I) Cardiac TG level. Results are expressed as the fold difference compared with *dSir2-*control flies. (J) Relative heart *dFAS* gene expression level. (K) Illustrating qualitative differences in heart function parameters (10 s): fractional shortening and arrhythmia index; K1: *dSir2*-control; K2: *dSir2*-KD-ND; K3: *dSir2*- KD-ND+E; K4: *dSir2*- KD-HFD; K5: *dSir2*-KD-HFD+E. Independent sample *t*-test was used to identify differences between *dSir2*-control flies and *dSir2*-KD flies. A two-way ANOVA was used to identify differences among the ND, ND+E, HFD, and HFD+E groups in *dSir2*-OE flies. Independent sample *t*-test was used to identify differences between *dSir2*-control flies and *dSir2*-KD flies. Data are represented as means±s.e.m. \**P*\<0.05; \*\**P*\<0.01. The sample size was the same as with *w^1118^*.
Although it had been confirmed that a HFD could induce lipotoxic cardiomyopathy in *w^1118^* flies, it remained unclear whether lipotoxic cardiomyopathy, induced by cardiac *dSir2* knockdown, could be aggravated after a HFD intervention. To figure this out, the cardiac *dSir2* gene knockdown flies were fed a HFD. Results showed that the cardiac *dSir2* expression level, dSIR2 level, NAD^+^ level and *PGC-1α* expression level of *dSir2*-KD-HFD flies were lower than that of *dSir2*-KD-ND flies (*P*\<0.05) ([Fig. 5](#BIO044719F5){ref-type="fig"}A--D). This suggested that a HFD could reduce cardiac *dSir2* gene expression and the activity of NAD^+^/dSIR2/*PGC-1α* pathway in untrained *dSir2*-KD flies. In addition, heart diastolic diameter and fractional shortening of *dSir2*-KD-HFD flies were lower than that of *dSir2*-KD-ND flies (*P*\<0.05, *P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}E,G), and the arrhythmia of *dSir2*-KD-HFD flies was higher than that of *dSir2*-KD-ND flies (*P*\<0.05) ([Fig. 5](#BIO044719F5){ref-type="fig"}H,K). It indicated that a HFD could weaken cardiac contractility and increase the risk of arrhythmia in untrained *dSir2*-KD flies. What is more, the heart TG level and *dFAS* expression of *dSir2*-KD-HFD flies were higher than that of *dSir2*-KD-ND flies (*P*\<0.05, *P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}I,J). It also indicated that a HFD could increase cardiac lipid accumulation in untrained *dSir2*-KD flies. So, these results confirmed that a HFD could aggravate lipotoxic cardiomyopathy induced by cardiac *dSir2* knockdown via inhibiting cardiac NAD^+^/dSIR2/*PGC-1α* pathway in untrained flies.
Although it had been identified that exercise could prevent lipotoxic cardiomyopathy induced by a HFD in *w^1118^* flies, and although exercise combined with overexpression of cardiac *dSir2* could better prevent lipotoxic cardiomyopathy in old flies, it remained unknown as to whether endurance exercise could improve lipotoxic cardiomyopathy induced by cardiac *dSir2* knockdown, and whether endurance exercise could prevent further deterioration of lipotoxic cardiomyopathy induced by a HFD in cardiac *dSir2* knockdown flies. To figure this out, the *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies were given exercise training. Results showed that endurance exercise significantly upregulated the expression of cardiac *dSir2* gene in both *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies (*P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}A), and it also remarkably increased heart dSIR2 level, NAD^+^ level and *PGC-1α* expression level in both *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies (*P*\<0.05 and *P*\<0.01, respectively) ([Fig. 5](#BIO044719F5){ref-type="fig"}B--D). In addition, endurance exercise significantly increased diastolic diameter and fractional shortening in both *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies (*P*\<0.05 and *P*\<0.01, respectively) ([Fig. 5](#BIO044719F5){ref-type="fig"}E,G), and it significantly decreased arrhythmia index in both *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies (*P*\<0.05 and *P*\<0.01, respectively) ([Fig. 5](#BIO044719F5){ref-type="fig"}H,K). Moreover, endurance exercise significantly reduced heart TG level and *dFAS* expression in both *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies (*P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}I,J). Therefore, we hypothesized that endurance exercise could improve lipotoxic cardiomyopathy induced by cardiac *dSir2* knockdown, and it could prevent further deterioration of lipotoxic cardiomyopathy induced by a HFD in cardiac *dSir2* knockdown flies.
However, to find out the extent to which exercise saved lipotoxic cardiomyopathy in both *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies, the *dSir2*-KD-HFD+E flies were compared with *dSir2*-KD-ND flies, and the *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies were compared with *dSir2*-control flies. Interestingly, the cardiac *dSir2* expression level, dSIR2 level, NAD^+^ level, *PGC-1α* expression level, diastolic diameter and fractional shortening of *dSir2*-KD-HFD+E flies were higher than that of *dSir2*-KD-ND flies (*P*\<0.05 and *P*\<0.01, respectively) ([Fig. 5](#BIO044719F5){ref-type="fig"}A--E,G). The arrhythmia index, heart TG level, and *dFAS* expression of *dSir2*-KD-HFD+E flies were lower than that of *dSir2*-KD-ND flies (*P*\<0.05, *P*\<0.01) ([Fig. 5](#BIO044719F5){ref-type="fig"}H--J). Moreover, the cardiac *dSir2* expression level, dSIR2 level, NAD^+^ level, *PGC-1α* expression level, diastolic diameter, fractional shortening, arrhythmia index, heart TG level and *dFAS* expression of *dSir2*-KD-ND flies and *dSir2*-KD-HFD flies were not significantly different from that of *dSir2*-KD-HFD flies (*P*\>0.05) ([Fig. 5](#BIO044719F5){ref-type="fig"}). So, we confirmed that endurance exercise could resist and treat lipotoxic cardiomyopathy induced by a HFD and cardiac *dSir2* knockdown flies.
DISCUSSION {#s3}
==========
Lipotoxic cardiomyopathy induced by a HFD related to inhibiting NAD^+^/dSIR2/*PGC-1α* pathway {#s3a}
---------------------------------------------------------------------------------------------
Increasing evidence showed that a HFD could induce lipotoxic cardiomyopathy, which manifested as cardiac lipid accumulation, reduced cardiac contractility, increased risk of arrhythmia and severe structural pathologies in flies ([@BIO044719C8]; [@BIO044719C19]; [@BIO044719C49]; [@BIO044719C70]). Similarly, our results confirmed again that a HFD could induce lipotoxic cardiomyopathy. For example, a HFD increased the heart TG level and the heart *dFAS* expression levels, and it resulted in heavy lipid accumulation. In addition, a HFD reduced heart fractional shortening via decreasing diastolic diameter, and it increased arrhythmia indexes. These changes easily led to heart dysfunction and heart failure ([@BIO044719C5]; [@BIO044719C34]; [@BIO044719C65]). Moreover, the heart NAD^+^ levels, dSIR2 levels, *dSir2* gene expression levels and *PGC-1α* expression were all decreased after feeding a HFD. These results suggested the heart NAD^+^/dSIR2/*PGC-1α* pathway activation was inhibited by a HFD.
Accumulating studies indicated that HFD-induced obesity could decrease the levels of NAD^+^ by several ways. For example, the mitochondria had compromised function due to overload that was thought to be induced by excessive beta-oxidation, and HFD-induced obesity could also decrease the mitochondrial numbers ([@BIO044719C2]; [@BIO044719C74]). Since there was a NAD^+^ pool in mitochondrial, NAD^+^ levels in metabolic tissues decreased with obesity ([@BIO044719C3]; Di Lisa et al., 2001; [@BIO044719C58]). In addition, the oxidative stress by a HFD also caused an increased in lipid peroxidation, and the increased oxidative stress accumulated fat was an important pathogenic mechanism of metabolic syndromes associated with obesity ([@BIO044719C17]; [@BIO044719C55]). A number of studies had demonstrated that oxidative stress induced excessive PARP-1 activation mediates cell death, and NAD^+^ depletion mediates PARP-1-induced cell death, which indicated that HFD-induced increased oxidative stress could also reduce NAD^+^ level via PARP-1 activation mediated cell death ([@BIO044719C29]; [@BIO044719C39]; [@BIO044719C46]; [@BIO044719C48]; [@BIO044719C51]). NAD^+^ depletion lead to abnormal hepatic lipid metabolism in HFD-induced nonalcoholic fatty liver disease ([@BIO044719C75]). Therefore, these reports hinted that a HFD induced decreased NAD^+^ levels in cells.
In HFD flies, because of the lipotoxicity the reduction of cardiac NAD^+^ content inhibited dSIR2 (NAD-dependent histone deacetylases) activity. Since *PGC-1α* was a key antagonist of HFD-induced lipotoxic cardiomyopathy, and since PGC-1α was activated by the NAD^+^-dependent deacetylate of SIR2 ([@BIO044719C4]; [@BIO044719C56]), HFD-induced the reduction of NAD^+^ content decreased PGC-1α activation via inhibiting dSIR2 activation. This finally led to a more severe mitochondria number reduction, lipid accumulation, and dysfunction in the heart of HFD-fed flies. These findings suggested that the activity of NAD^+^/dSIR2/*PGC-1α* pathway was closely related to the formation of lipotoxic cardiomyopathy in flies. However, to confirm the relationship between NAD^+^/dSIR2/*PGC-1α* pathway and lipotoxic cardiomyopathy, further experiments should be done.
Cardiac *dSir2* gene involved into regulating the formation of lipotoxic cardiomyopathy {#s3b}
---------------------------------------------------------------------------------------
It had been proposed that *Sir2* in the fat body plays an important role in regulating fat storage and mobilization, as *Sir2/Sirt1* had been implicated in regulation of fat metabolism in flies and mammals. Overexpression of *Sir2* in the adult fat body was found to be sufficient to extend the lifespan of male and female *Drosophila* ([@BIO044719C10]; [@BIO044719C28]; [@BIO044719C36]). However, it remains unclear whether *Sir2* can regulate heart lipid metabolism and cardiac function.
As expected, our research showed that the *dSir2* gene could also regulate lipid metabolism in the heart. For example, overexpression of cardiac *dSir2* gene could not only reduce heart TG accumulation and heart *dFAS* expression, but also it could prevent heart TG accumulation and decrease heart *dFAS* expression induced by a HFD. While cardiac *dSir2* knockdown could increase heart TG accumulation and heart *dFAS* expression, and this would get worse under a HFD intervention. The *dSir2* apparently regulates expression of genes involved in fat metabolism, and lack of *Sir2* increases fat deposition under normal conditions and consequently impairs starvation survival capabilities of flies ([@BIO044719C7]). It has been reported that the FOXO and PGC-1*α* activity can be modulated by SIR2 deacetylation ([@BIO044719C16]; [@BIO044719C41]), and these two factors are important to the formation of lipotoxic cardiomyopathy in *Drosophila* ([@BIO044719C19]; [@BIO044719C20]). However, there is no direct evidence that cardiac *Sir2* can regulate heart lipid metabolism via activating PGC-1*α* activity in *Drosophila*.
In this study, we also found that cardiac *dSir2* gene had the ability to modulate cardiac function. For instance, overexpression of cardiac *dSir2* gene could increase heart fractional shortening via increasing diastolic diameter, and overexpression of cardiac *dSir2* gene could decrease arrhythmia index. In addition, overexpression of cardiac *dSir2* gene could prevent fractional shortening decline and arrhythmia index increase induced by a HFD. Also, cardiac *dSir2* knockdown could decrease heart fractional shortening via decreasing diastolic diameter, and cardiac *dSir2* knockdown could increase arrhythmia index. A HFD intervention could aggravate cardiac systolic dysfunction and arrhythmia in cardiac *dSir2* knockdown flies. It has been reported that *FOXO* or *PGC-1α* knockdown results in serious cardiac dysfunction in flies, including reduced cardiac contractility and increased the risk of arrhythmia ([@BIO044719C19]; [@BIO044719C20]). The activity of PGC-1α can be modulated by SIR2 deacetylation ([@BIO044719C16]; [@BIO044719C41]). However, there was no direct evidence that cardiac *dSir2* could regulate heart function via activating PGC-1*α* activity in *Drosophila*.
To confirm whether cardiac *dSir2* gene could regulate lipotoxic cardiomyopathy by modulating cardiac NAD^+^/ dSIR2/*PGC-1α* pathway, the cardiac dSIR2 level, NAD^+^ level and *PGC-1α* expression levels were measured. We found overexpression of cardiac *dSir2* gene increased cardiac dSIR2 level, NAD^+^ level, and *PGC-1α* expression level, and overexpression of cardiac *dSir2* gene could prevent the decline of cardiac dSIR2 level, NAD^+^ level, and *PGC-1α* expression level induced by a HFD. On the other hand, cardiac *dSir2* knockdown decreased cardiac dSIR2 level, NAD^+^ level, and *PGC-1α* expression level. A HFD intervention could aggravate the decline of cardiac dSIR2 level, NAD^+^ level, and *PGC-1α* expression level in cardiac *dSir2* knockdown flies.
SIR2 directly binds to PGC-1α and deacetylated it in 293T cells and PC12 cells ([@BIO044719C43]). SIR2 stimulates the ability of PGC-1α to coactivate hepatocyte nuclear factor 4α, thereby positively regulating gluconeogenic genes in response to pyruvate in hepatic cells. In the same cell type, SIR2 also enhances the ability of PGC-1α to inhibit glycolytic genes in response to pyruvate ([@BIO044719C73]). Furthermore, SIR2 affects fatty acid oxidation in adipocytes and knockdown of PGC-1α cancels the effect of overexpression of SIR2 upon fatty acid oxidation. Thus the availability of other family members also contributes to the net effect of sirtuins upon PGC-1α ([@BIO044719C62]). Systemic deletion of SIR2 in mice induces the development of dilated cardiomyopathy, which is accompanied by mitochondrial dysfunction. Overexpression of SIR2 in pancreatic β-cells enhances insulin secretion in response to glucose and improves glucose metabolism by increasing ATP production via suppression of uncoupling protein-2 expression ([@BIO044719C13]; [@BIO044719C45]).
In this research, we identified that cardiac *dSir2* gene could regulate *PGC-1α* expression via NAD^+^ and dSIR2 level. This possibly caused dysfunction of the heart\'s mitochondria, and this could induce lipid accumulation and reduced contractile ability of the heart, and eventually result in lipotoxic cardiomyopathy. A previous study has confirmed that when *PGC-1* was overexpressed in heart, lipotoxic cardiomyopathy was prevented after flies were fed a HFD ([@BIO044719C19]). Similarly, overexpression of *dSir2* also resisted lipotoxic cardiomyopathy induced by a HFD in this study. Therefore, based on this evidence, we declared that the activation of cardiac NAD^+^/dSIR2/*PGC-1α* pathway was a key pathway that regulated the formation of lipotoxic cardiomyopathy.
Exercise improved lipotoxic cardiomyopathy induced by a HFD and *dSir2* knockdown in old *Drosophila* {#s3c}
-----------------------------------------------------------------------------------------------------
A lot of studies have confirmed that appropriate endurance exercise is a healthy and economical way to prevent and cure obesity, and endurance exercise is also considered a good way to improve heart functional in obese or old individuals ([@BIO044719C61]). For example, exercise training can strengthen the heart\'s ability to use fatty acids to provide energy by increasing the activity of related enzymes, which prevents lipid excessive accumulation in the heart ([@BIO044719C67]). Furthermore, exercise training improves heart function such as cardiac contractibility and exercise reduces heart failure in obese individuals ([@BIO044719C24]; [@BIO044719C40]; May et al., 2016; [@BIO044719C66]). Finally, exercise increases muscle NAD^+^ levels and neuron NAD^+^ levels, and it activates transcriptional activity of *PGC-1α* and increases mitochondrial density ([@BIO044719C16]; [@BIO044719C25]; [@BIO044719C38]; [@BIO044719C41]). However, it remains unclear whether exercise can prevent lipotoxic cardiomyopathy by activating cardiac NAD^+^/dSIR2/*PGC-1α* pathway.
In this research, we found that in HFD-fed flies, the heart TG level and the heart *dFAS* expression level were significantly decreased after exercise training. In addition, exercise training increased heart fractional shortening via increasing diastolic diameters, and exercise training decreased arrhythmia index in a HFD-fed flies and cardiac *dSir2* knockdown flies. Finally, the heart NAD^+^ levels, dSIR2 levels, *dSir2* gene expression levels and *PGC-1α* expression were all increased after exercise training in HFD-fed flies. This suggested that exercise training reduced lipid accumulation, improved heart function, activated NAD^+^/dSIR2/*PGC-1α* pathway and reduced the risk of arrhythmia, which prevented lipotoxic cardiomyopathy formation.
In HFD-fed flies, on one hand, exercise training can strengthen the heart\'s ability to use fatty acids to provide energy by increasing the activity of related enzymes, which prevents excessive lipid accumulation in the heart ([@BIO044719C67]). On the other hand, because the dSIR2 protein plays a pivotal role in PGC-1α function via NAD-dependent deacetylation ([@BIO044719C30]; [@BIO044719C33]; [@BIO044719C37]; [@BIO044719C54]), and because the *PGC-1α* is a key antagonist of HFD-induced lipotoxic cardiomyopathy ([@BIO044719C8]; [@BIO044719C26]; [@BIO044719C52]; [@BIO044719C71]), exercise training increased the *PGC-1α* function via improving heart NAD^+^ content and heart dSIR2 activation in this study ([@BIO044719C16]; [@BIO044719C25]; [@BIO044719C38]; [@BIO044719C41]). Therefore, these results suggested that the NAD^+^/dSIR2/*PGC-1α* pathway activation was an important molecular mechanism of exercise resistance against lipotoxic cardiomyopathy. However, it remained unclear whether exercise training could improve lipotoxic cardiomyopathy induced by cardiac *dSir2* knockdown.
In this study, the heart *dSir2*-RNAi flies were exercise trained and we found exercise training also reduced lipid accumulation, enhanced heart function, activated NAD^+^/dSIR2/*PGC-1α* pathway, and reduced the risk of arrhythmia, which improved lipotoxic cardiomyopathy induced by heart *dSir2* RNAi. In heart *dSir2*-RNAi flies, exercise training can also increase the cardiac ability to use fatty acids to provide energy by increasing the activity of related enzymes, which prevents lipid excessive accumulation in the heart ([@BIO044719C67]). In addition, increasing evidence hints that exercise training can upregulate *dSir2* activity. For instance, it has been reported that exercise not only improves blood NAD^+^ levels but also in muscle and cardiac NAD^+^ levels ([@BIO044719C12]; [@BIO044719C22]; [@BIO044719C64]), which may eventually result in increasing NAD^+^ activity in these tissues and organs to meet the demand of NAD^+^ metabolism during exercise training. Since the *dSir2* activity can be regulated by free NAD^+^ in cells, the *dSir2* expression may be indirectly elevated by exercise trained ([@BIO044719C23]). Moreover, increasing evidence indicates that exercise can intensify the contraction of cardiac muscles which may facilitate SIRT1 upregulation. Exercise training can upregulate heart AMPK expression -- AMPK is an energy sensor. Since AMPK also increases the intracellular NAD^+^ levels, its activity is correlated with SIRT1 enhancement ([@BIO044719C14]; [@BIO044719C42]). Next, recent studies report that exercise training upregulates SIRT1 in kidney, liver and brain ([@BIO044719C31]; [@BIO044719C44]). So, these reports suggest that exercise training can increase *Sir2* expression. However, there is no evidence that exercise training can increase heart *dSir2* expression by affecting the UAS/Gal4 system. Besides, in our experiment, a stronger cardiac *dSir2* knockdown has not been generated any other way, such as with an inducible CRISPR or a stronger RNAi line or stronger driver. Therefore, our results indicated that exercise training rescued the cardiac *dSir2* expression and dSir2 protein levels only under this mild *dSir2*-knockdown condition, and the reason may be that exercise induction of cardiac *dSir2* was stronger than knockdown.
In conclusion, we identified that the heart *dSir2* gene and *dSir2*/NAD^+^/*PGC-1α* pathway regulated the heart lipid metabolism and the formation of lipotoxic cardiomyopathy. Lipotoxic cardiomyopathy could be induced by heart *dSir2* knockdown, but the heart *dSir2* overexpression could prevent a HFD-induced lipotoxic cardiomyopathy. Exercise training could improve lipotoxic cardiomyopathy induced by a HFD or heart *dSir2* knockdown in old *Drosophila*. The NAD^+^/dSIR2/*PGC-1α* pathway activation was an important molecular mechanism of exercise resistance against lipotoxic cardiomyopathy.
MATERIALS AND METHODS {#s4}
=====================
Fly stocks, diet and husbandry {#s4a}
------------------------------
The *w^1118^* and *hand-Gal4* line was a gift from Xiu-shan Wu (Heart Development Center of Hunan Normal University). UAS-*dSir2*-OE (*w^1118^;P{EP}Sirt1^EP2300^DnaJ-H^EP2300^/CyO*) line was obtained from the Bloomington Stock Center, and the P{EP} construct carries Scer\\UAS binding sites for the Scer\\GAL4 transcriptional regulator, and bacterial sequences that allow plasmid rescue. The Gal4-UAS system allows regulated expression of genes proximate to the site of the insertion: genes properly oriented with respect to the Scer\\UAS sequences can be conditionally expressed via transgene-derived Scer\\GAL4 activity ([@BIO044719C57]). UAS-*dSir2*-KD (*w^1118^; P{GD11580}v23201*) line was obtained from the Vienna Drosophila RNAi Center. Male *hand-Gal4* flies were crossed to female UAS-*dSir2*-OE flies and UAS-*dSir2*-KD flies.
To avoid the influence of genetic background differences on the results, maternal origin was used as the genetic control. The female '*w^1118^;P{EP}Sirt1^EP2300^DnaJ-H^EP2300^/CyO*' and '*hand-Gal4\>w^1118^;P{EP}Sirt1^EP2300^DnaJ-H^EP2300^/CyO*' were represented as '*dSir2*-control' and '*dSir2*-OE'. The female '*w^1118^; P{GD11580} v23201*' and '*hand-Gal4\>w^1118^; P{GD11580} v23201*' were represented as '*dSir2*-control' and '*dSir2*-KD'. The female '*w^1118^*', '*dSir2*-OE', and '*dSir2*-KD' virgin flies were divided into several groups: normal-diet (ND) group, normal-diet +exercise (ND+E) group, high-fat diet (HFD) group, and HFD+exercise (HFD+E) group.
Normal food contained 10% yeast, 10% sucrose and 2% agar. The HFD was made by mixing 30% coconut oil with the food in a weight to volume ratio with the normal food ([@BIO044719C8]). Both HFD+E group flies and HFD-E group flies were fed the HFD from 28 days of age and were exposed to the HFD for 5 consecutive days. During the experimental time course, flies were housed in a 22±1°C incubator with 50% humidity and a 12-h light/dark cycle. This environment could keep the coconut oil food in a solid state since the melting point of coconut oil is about 24°C, thus ensuring that flies would not get stuck in the oily food. Fresh food was provided every other day for the duration of the experiment. All group flies were raised to the fourth weekend. Flies were trained or fed a HFD at 5 weeks old as we found flies were very sensitive to exercise or HFDs at this time.
Exercise training device and protocols {#s4b}
--------------------------------------
The advantage of the flies\' natural negative geotaxis behavior was taken to induce upward walking when constructing the exercise device ([@BIO044719C63]). All exercise group flies started exercise from when they were 5 weeks old, and underwent a 5-day-long exercise program. Vials were loaded horizontally into a steel tube that was rotated about its horizontal axis by an electric motor, with a gear regulating its shaft speed. There were 25 flies in each vial. Thus, each vial was rotated along its long axis with the accompanying rotating steel tube, which made the flies climb. Most flies continued to respond by climbing throughout the exercise period. The few that failed to climb were actively walking at the inner wall of the vial ([@BIO044719C69]; [@BIO044719C76]). Flies were exercised in vials with a 2.8-cm inner diameter, rotated at 0.14 rev/s. Flies were exercised for 1.5 hours every time.
Semi-intact *Drosophila* preparation and image analysis {#s4c}
-------------------------------------------------------
First, 30 flies were anesthetized with FlyNap for 2--3 min (a few flies were anesthetized with FlyNap for 4--5 min as they were hard to narcotize). Next, the head, ventral thorax and ventral abdominal cuticle were removed by special scissors and tweezers to expose the heart in the field of vision of a microscope. Note, dissections were done under oxygenated artificial hemolymph. These semi-intact preparations were allowed to equilibrate with oxygenation for 15--20 min before filming. Finally, image analysis of heart contractions was performed using high-speed videos of the preparations. Videos were taken 120--130 frames per second using a Hamamatsu EM-CCD digital camera on a Leica DM LFSA microscope with a 10 immersion lens. To get a random sampling of heart function, a single 30-s recording was made for each fly. All images were acquired and contrast enhanced by using Simple PCI imaging software. The heart physiology of the flies was assessed using a semi-automated optical heartbeat analysis program that quantifies heart diastolic diameters, systolic diameters, fractional shortening and arrhythmia index ([@BIO044719C21]).
The SIRT1/dSIR2 assay, NAD^+^ assay, and TG assay {#s4d}
-------------------------------------------------
For dSIR2 assay, 80 hearts were homogenized in 200 μl PBS buffer. To break the cells, hearts were subjected to freeze-thaw cycles. The homogenates were then centrifuged for 5 min at 5000×***g*** to get the supernate as a sample. We used purified insect SIRT1 antibody to coat microtiter plate wells, made the solid-phase antibody, then added SIRT1 to the wells, which, combined with HRP-labelled antibody, become antibody-antigen-enzyme-antibody complex. Then we added 50 μl standard sample liquid and experimental sample liquid to different wells. 100μl of HRP-conjugate reagent was added to each well, which were then incubated for 60 min at 37°C. 50 μl Chromogen Solution A and Chromogen Solution B were added to each well, which were kept in the dark for 15 min at 37°C. Next 50 μl of Stop Solution was added to each well to stop the reaction (a color change from blue to yellow was observed). The blank well was taken as zero, absorbance was read at 450 nm after adding Stop Solution and within 15 min. The OD values were used to determine sample Sirt1 concentration from the standard curve according to the manufacturer\'s instructions (Insect SIRT1 ELISA Kit, MLBIO).
For NAD^+^ assay, 80 hearts were homogenized in 200 μl NAD exaction buffer, and then extracts were heated at 70°C for 5 min. 20 μl assay buffer and 100 μl of the opposite extraction to neutralize the extracts were added. Samples were centrifuged at 14,000 rpm for 5 min and then 40 μl standard sample liquid and experimental sample liquid were transferred to separate wells. 80 μl of working reagent (40 μl assay buffer, 1 μl oxidation-reduction enzyme, 10 μl 10% ethanol, 20 μl PMS, 20 μl MTT) was quickly added to each well. The resulting samples were then measured at 565 nm with a microplate spectrophotometer. The OD values were used to determine sample NAD^+^ concentration from the standard curve according to the manufacturer\'s instructions (EnzyChrom™NAD^+^/NADH assay kit).
Cardiac TG measurements were taken from the hearts of 20 female flies dissected in artificial hemolymph. Care was taken to remove as much adipose tissue and other heart-associated cells from the heart as possible. Preparations were washed two times with PBS. Using fine forceps, the hearts were pulled off from the cuticle and transferred into Eppendorf tubes containing 26 μl of PBST (PBS+0.05% Triton X-100). Tubes were immediately frozen and stored at −80°C. Hearts were then mildly sonicated to lyse the cells. 20 μl heart lysates was transferred to a 96-well plate containing 200 μl of lipid reagent. After incubating 10′ at 37°C, 3 μl of heart lysate was transferred to wells containing 200 μl of Bradford reagent to measure protein content. The reaction mixture was incubated in an Environ Shaker at 300 rpm at 37°C for 10 min; the OD 550 nm was measured using SpectraMax Molecular Devices Corp and compared with a standardized curve. Experiments were repeated at least three times in multiple replicates.
qRT-PCR {#s4e}
-------
To check the transcriptional expression of the *dSir2*, *dFAS* and *PGC-1α* gene, 80 flies\' hearts were homogenized in Trizol for each group. Firstly, 10 μg of the total RNA was purified by organic solvent extraction from the Trizol (TRIzol, Invitrogen). Next, the purified RNA was treated with DNase I (RNase-free, Roche) and used to produce oligo dT-primed cDNAs (SuperScript II RT, Invitrogen), which were then used as templates for quantitative real-time PCR. The rp49 gene was used as an internal reference for normalizing the quantity of total RNAs. Finally, Real-time PCR was performed with SYBR green using an ABI7300 Real time PCR Instrument (Applied Biosystems). Expression of the various genes was determined by the comparative CT method (ABI Prism 7700 Sequence Detection System User Bulletin \#2, Applied Biosystems). Primer sequences of *dSir2* were as follows: F: 5′-GCAGTGCCAGCCCAATAA-3′; R: 5′-AGCCGATCACGATCAGTAGA-3′. Primer sequences of *PGC-1α* were as follows: F: 5′-TGTTGCTG CTACTGCTGCTT-3′; R: 5′-GCCTCTGCATCACCTACA CA-3′. Primer sequences of *dFAS* were as follows: F: 5′-GGTGAGACCATCGTGGAAGT-3′; R: 5′-AATGTCTGCCAAGCCAGAGT-3′. Primer sequences of *dnaJ-H* were as follows: F: 5′-GCAAGATGGCACACGTAGCTG-3′; R: 5′-CCACTGTAGCAAC ACGTAATCACC-3′. Primer sequences of Internal were as follows: F: 5′-CTAAGCTGTC GCACAAATGG-3′; R: 5′-AA CTTCTTG AATCCGGTGGG-3′.
Statistical analyses {#s4f}
--------------------
A two-way ANOVA was used to identify differences among the ND, ND+E, HFD, and HFD+E groups of *Drosophila* with the same genetic background. Independent sample *t*-test was used to identify differences between *dSir2*-control flies and *dSir2*-OE flies. Analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 16.0 for Windows (SPSS Inc., Chicago, USA), with statistical significance set at *P*\<0.05. Data are represented as means±s.e.m.
The authors thank Xiu-shan Wu (The Center for Heart Development, College of Life Science, Hunan Normal University) for supporting *Drosophila* of *w^1118^* and heart Shoot software technology. We also thank Karen Ocorr and Rolf Bodmer (Neuroscience and Aging Research Center, Sanford Burnham Medical Research Institute) for supporting semi-automatic optical echocardiography analysis software.
**Competing interests**
The authors declare no competing or financial interests.
**Author contributions**
Methodology: D.-T.W., D.C., Y.L.; Software: L.Z., Y.L.; Formal analysis: J.-x.L., K.L., W.-q.H.; Resources: D.-T.W., L.Z.; Data curation: J.-x.L., K.L., W.H.; Writing - original draft: D.-T.W.; Funding acquisition: L.Z.
**Funding**
This work was supported by the National Natural Science Foundation of China \[No. 31671243\] and the Ministry of Education funding PhD programs \[No. 20134306110009\].
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-sensors-18-03497}
===============
The world's seas are a valuable resource as well as a key element in its ecology, and are in need of protection as an important source of food, wealth and life. For this, systems and technologies are needed to monitor their status and guarantee their sustainable management. The successful management of marine resources involves monitoring the physical and chemical parameters related to water quality, such as salinity, temperature, dissolved oxygen, nitrates, density, and chlorophyll levels, among others. Other reasons for monitoring the seabed include, detecting and preserving artefacts of archaeological value, monitoring the status of marine flora and fauna, especially sensitive species in danger of extinction, and detecting and removing man-made contaminants and plastics. This last point is particularly important as over time plastic fragments into microplastics, which are toxic to both marine life and humans. According in \[[@B1-sensors-18-03497]\], in 2050 there will be more tons of plastic waste than fish in the oceans. The EU has declared war on plastic waste and is preparing a recycling strategy in \[[@B2-sensors-18-03497]\] to ensure that every piece of packaging on the continent is reusable or recyclable by 2030, with an investment of €350 M over the next few years.
Although oceanographic observatories can monitor large expanses of water continuously and efficiently, and are normally based on static buoys or wireless or wired underwater sensor networks, they can only take measurements at fixed points. When large areas have to be covered specific operations have to be planned and executed. The cost, in terms of time and money, of such operations is normally high and in proportion to the area monitored. The deployment of a large number of buoys is unfeasible due to their impact on the environment and their vulnerability to vandalism. To obtain trustworthy readings and avoid errors when extrapolating measurements to other areas it is fundamental to employ appropriate purpose-built monitoring systems.
Marine robots, like Remotely Operated Vehicles (ROVs), Autonomous Underwater Vehicles (AUVs), or Autonomous Surface Vehicles (ASVs), do not suffer from the main limitations of buoys and are now available for a wide range of observation missions. However, they all have limited autonomy in terms of both time and the area they can cover. In addition, they must be transported to the observation area, deployed for a certain time and picked up when the mission is over, all of which involve extra costs.
The power issue is one of the fundamental challenges of increasing AUV mission times. Renewable energy sources must be used because fuel-based solutions will run out of fuel sooner or later. The current photovoltaic solar systems can generate around 1754 W/m^2^ in good weather conditions. With the limited ASVs recharging surface area, this energy is not enough to keep the thrusters and control systems in permanent operation on long missions and recharging operations must be as efficient as possible to minimise standby times. These operations can be difficult in bad weather conditions, may require more power to avoid drifting than what is effectively injected into the batteries, and may even result in a system shutdown.
From the detailed review of the state of the art in [Section 2](#sec2-sensors-18-03497){ref-type="sec"}, we conclude that no solution involving autonomous vehicles has yet been proposed that guarantees a permanent presence in the exploration area. Given the limitations of both types of existing solutions (buoys and autonomous marine vehicles) in this paper we propose a novel approach to explore shallow waters, which we call the BUSCAMOS Robotic Observatory (BUSCAMOS-RobObs). The system is a combination of autonomous vehicles and fixed buoys, whose complete power and navigation autonomy is achieved thanks to a fuzzy decision-making software architecture.
The proposed solution has an ASV's freedom of movement and can anchor itself to the seabed and become a "buoy", either for measuring purposes or recharging the batteries, thus ensuring the energy autonomy required for carrying out permanent operations. The vehicle thus avoids drifting and possible accidents in the "buoy mode", especially important near the coast, and resumes monitoring tasks when the required power levels are reached, all of its operations being managed by the vehicle's central control and decision-making system. The decision-making system calculates the route to the next area to be explored taking into account a number of parameters, including: vehicle position, the water's physical and chemical parameters, distance to the next exploration area, solar radiation, available energy, wind speed and direction, water currents, etc.
This article describes the hardware and software innovations included in BUSCAMOS-RobObs. To demonstrate the feasibility of the proposed solution, we include the crucial preliminary simulations of power and decision-making autonomy, together with descriptions of actual observation missions and their results in the Mar Menor (see Figure 6 in [Section 4](#sec4-sensors-18-03497){ref-type="sec"}) in southeast Spain. This article is organized as follows: [Section 2](#sec2-sensors-18-03497){ref-type="sec"} outlines the current state of the art and summarises the vehicle's technical specifications, which is described in greater detail in [Section 3](#sec3-sensors-18-03497){ref-type="sec"} (hardware and software, power generation and energy management). The design of the strategic decision-making layer and the application to the observation mission is outlined in [Section 4](#sec4-sensors-18-03497){ref-type="sec"}. [Section 5](#sec5-sensors-18-03497){ref-type="sec"} gives details of the tests and their results, while the main conclusions are shown in [Section 6](#sec6-sensors-18-03497){ref-type="sec"}.
2. State of the Art {#sec2-sensors-18-03497}
===================
This section describes the state of the art of existing technologies that led us to propose a dual approach (buoy-ASV) to a mobile robotic observatory. The two main types of solutions that already exist to carry out marine monitoring are analysed: static systems (sensor networks and buoys) and mobile systems based on autonomous vehicles. Given the aim of designing a vehicle capable of permanent presence in a marine environment, the present state of the art of the methods proposed to improve the energy autonomy of autonomous systems is given special attention.
2.1. Static Monitoring Solutions: Buoys and Sensor Networks {#sec2dot1-sensors-18-03497}
-----------------------------------------------------------
According to \[[@B3-sensors-18-03497]\], oceanographic observatories measure physical, chemical, and biological variables in coastal waters, perform real-time observations and supply information to help in decision-making on managing problems such as climate change, the human impact, coastal variability, natural disasters, or the condition of ecosystems. They normally consist of fixed buoys with different sensors to measure the parameters of interest. The present authors advocate using Wireless Sensor Networks (WSN) as an innovative solution for the deployment of multiple buoys in coastal areas, as they are cost-effective and have low energy consumption (powered by solar panels), which allows a higher number of sensors to be deployed with different network topologies and protocols at a reasonable cost in shallow coastal waters.
The review of wireless sensor networks in marine environment monitoring described in \[[@B4-sensors-18-03497]\] makes interesting comparisons of the leading current marine wireless sensor networks and conclude that when a buoy-based system is used for long-term observation the sensors are susceptible to bio-fouling (microbial and algal films). Other problems highlighted in these articles include the need for flotation and mooring devices, possible vandalism, interference from maritime traffic, the movement of nodes by tides, waves, vessels, etc., and the fact that an anchor is required to fix them to the seabed. Energy systems for sensor nodes generally include batteries, capacitors, thermal engines, fuel cells and solar energy collectors. The photovoltaic system is the most frequently used energy system at present in \[[@B5-sensors-18-03497]\]. However, using batteries to store energy has many limitations, such as a restricted energy capacity, the low energy levels obtained from solar panels or excessive consumption in \[[@B6-sensors-18-03497]\]. When the buoys run out of energy they stop transmitting. Exhausted batteries made of toxic heavy metals are a highly contaminating waste. Rechargeable batteries with an energy capacity greater than the daily energy consumed by the system are a better option and enable the power system to supply energy at night and in bad weather conditions in \[[@B7-sensors-18-03497]\].
Most of the aforementioned papers provide little information on power systems in general, and the energy management algorithms, if they exist, are limited to considering battery voltage and intensity and maybe the power obtained from the solar panels. In most cases small panels (2.5 W/8 V) are installed with 3.7 V/5 Ah batteries, as in \[[@B4-sensors-18-03497]\]. Some authors monitored the energy consumption of their electronic and electrical equipment in \[[@B8-sensors-18-03497]\], but did not use the information to create an intelligent control system for the diverse power requirements.
It should be noted that all these systems have a serious limitation: a deployed buoy can only provide values from its fixed measuring point. A number of buoys will be needed to cover a wider area with costs rising in proportion to the area monitored, unfeasible due to the impact on maritime traffic, the environment, and tourism.
2.2. Autonomous Monitoring Solutions: Marine and Underwater Vehicles {#sec2dot2-sensors-18-03497}
--------------------------------------------------------------------
Due to the intrinsic complexity of the marine environment, monitoring tasks in the open sea started with remotely operated vehicle (ROVs) and continued with the development of autonomous surface and underwater vehicles (ASVs and AUVs), which overcome the ROVs' main limitation: the need for a support vessel and a human controller. AUVs thus opened up new opportunities for sea monitoring and explorations as they could gather data autonomously for extended periods of time at a lower cost than ROVs.
To make the most of AUVs they must be able to carry out long missions, an important goal of research in autonomous systems in general, especially in marine robots (see for example the Special Issue on "Long-Term Autonomy in Marine Robotics" in \[[@B9-sensors-18-03497]\]). According to \[[@B10-sensors-18-03497]\], long-term autonomy can be defined as "the ability of a robot to operate robustly for an extended period of time (hours, days or even weeks), with reduced human supervision and in a real environment". Long-term autonomy can be considered from three points of view \[[@B11-sensors-18-03497]\], all related to the problems of operating robots over long periods of time: (i) changing state of the sea and the perishability of current knowledge; (ii) unforeseen hardware and software failures during extended operations; and (iii) unexpected changes in mission goals, or new constraints and priorities for the robot to execute, which can also cause conflicts.
Achieving any of the aforementioned levels of autonomy requires first of all achieving the maximum possible power autonomy. This section describes the main initiatives in the literature to develop marine observation tasks, with special emphasis on the energy management system used for long-term operations at sea.
Gliders are, perhaps, the best example of marine vehicles in terms of energy autonomy. They use hardly any energy to move and by moving their wings and modifying their buoyancy they can convert vertical immersion movements into a horizontal movement, thus obtaining a very low power consumption. Although they are not as fast as other AUVs, gliders have a greater time and distance range than propeller-powered vehicles. Their measurement missions can last for months and cover thousands of kilometres. The so-called Wave Gliders in \[[@B12-sensors-18-03497]\] use the force of the waves to advance and do not use any thrusters. They consist of two parts joined by a cable: a surfboard floating on the surface and a second part about 8 m under the water. This system also uses solar panels to power the on-board electronics.
In general, Glider-type solutions are not suitable for exhaustive explorations of the marine environment, as these vehicles are at the mercy of currents and waves. Since they require little power to operate they have less accuracy regarding the area to be monitored. A system described in \[[@B13-sensors-18-03497]\] proposes a solution to this problem: an optimal mission-planning algorithm for a fleet of gliders based on the sampling on-demand paradigm. In this strategy, the user quantitatively sets the requirements related to the uncertainty over an area of interest that needs to be achieved by means of measurements taken by observing assets. In general, the uncertainty can vary from one region to another in the study area, according to the scientific/operational requirements of the mission. For instance, some areas may be considered more important to explore than others.
The more precise solutions, which provide better control of the areas explored, require the use of thrusters, and thrusters require a complete energy management strategy to ensure long periods of AUV activity. The remainder of this section summarises some of the AUVs described in the literature in order of decreasing autonomy. For each vehicle, we describe, when possible, its purpose, sensors and on-board electronics, measured mission time, energy source/s available and their management, and whether the system software takes into account energy issues when planning the missions.
SAUV II in \[[@B14-sensors-18-03497]\] is an exclusively solar-powered AUV, relatively small in size, with a total solar panel surface of 1 m^2^. The vehicle has a lithium battery (32 V, 2 kWh) as its energy storage system, which stores energy in the daytime to execute the missions at night. SAUV was designed to carry out long-endurance missions, such as monitoring, surveillance, or station-keeping. The longest mission recorded lasted for 30 days with 8 h of continuous operations.
A catamaran for environmental exploration in Rhode Island is described in \[[@B15-sensors-18-03497]\]. Given its small size, it can venture close to the shore. It has a few sensors to capture the physical-chemical parameters of the water and carries a radar sensor, GPS, altimeter for detecting the seabed and is controlled by two CPU based on Linux and a Windows operating system. The energy system is composed of a 5.5 kW diesel generator, which accumulates the energy in 4 batteries 12 V, which provide the energy to 48 V electric motors. Navigation time is limited to about 21 days.
The USV described in \[[@B16-sensors-18-03497]\] was designed and built to perform extended missions for environmental research by using an efficient route planner with Voronoi diagrams and a Dijkstra search algorithm. However, this algorithm does not take into account the solar energy generated to plan the most efficient route. The vessel is equipped with a GPS, AIS transponder, radar reflectors, depth sounder and navigation lights. Maximum navigation time is 3 months, powered by 1200 W solar panels, a wind turbine system with a power output of 720 W, and a 2.5 kW diesel generator system to recharge the batteries.
The catamaran described in reference \[[@B17-sensors-18-03497]\] has two electric motors powered by a mere 300 Wp solar module. Maximum navigation time is about 24 h. The ASV is equipped with a set of navigation sensors that includes a GPS, compass, depth sensor, laser scanner, camera, optical methane detector (OMD), a YSI Sonde (measures temperature, conductivity, chlorophyll, turbidity, dissolved oxygen, incident radiation), wind sensor and a profiled sonar, and laser-based obstacle-avoidance sensor. A profiling arm allows measurements of up to 5.5 m while the vehicle is moving.
Another vehicle used for shallow water and bathymetric studies includes a purpose-built software with integrated navigation and data acquisition systems in \[[@B18-sensors-18-03497]\]. This autonomous vessel has a navigation time of only two to four hours. As sensors, it carries a GPS, surface and submerged camera for video acquisition, ultrasonic obstacle detection, temperature control system, single beam echo sounder, inertial platform, an OLinuXino microcomputer, with a Linux operating system.
The vehicle described in \[[@B12-sensors-18-03497]\] also has a reduced autonomy (up to eight hours) although it is equipped with high-density rechargeable batteries. It carries sensors such as GPS, special electrodes for measuring anomalies in the distribution of electrical resistance along a channel, and temperature and oxygen sensors. According to the authors, it can be used to detect and scare birds away from drinkable water basins and fish-breeding ponds. The control system is based on a Windows OS.
The present authors gained previous experience in developing autonomous marine vehicles. In \[[@B19-sensors-18-03497]\] we describe a multi-vehicle system based on an ASV-UUV combination for oil-spill monitoring, "BUSCAMOS-OIL". One of its outstanding features is the incorporation of new control strategies based on bio-inspired neural networks to give adaptability and robustness to the ASV and UUV. The two vehicles are connected by an umbilical cable, which allows them to share power and computing resources. With its two robots, the platform has the ability to monitor large tracts of sea both on the surface and in the water column. With its time-series measurements, the system draws precise maps of the oil plume, with information on spill location, size, extent, direction and speed. To ensure its power supply, the ASV contains six photovoltaic panels of 130 Wp each, which recharge two 28 V batteries. It also has a diesel generator when no solar energy is available. This platform has now been improved for carrying out the missions described here, which require greater autonomy. The diesel generator has been eliminated, two more solar panels have been added and the explorer submarine has been removed, thus reducing energy consumption, weight and water friction. The vessel now generates more solar energy, which allows permanent navigation and exploration missions.
Besides stand-alone AUVs and ASVs, there are also other approaches that employ fleets of AUVs/ASVs to monitor and explore the seas. Each vehicle individually manages its energy sources, but as a fleet of vehicles these solutions provide extended capabilities to carry out the aforementioned missions. A fleet of ASVs for assessing greenhouse gas emissions from underwater plants in inland reservoirs, rivers and marshes is described in \[[@B20-sensors-18-03497]\]. Each ASV has certain energy limitations: it uses two 40 W solar modules, stores its energy in 12 V, 20 Ah lithium iron phosphate (LFP) batteries. Although energy management algorithms are not specifically considered in this article, it uses a navigation algorithm in which all the ASVs communicate with each other. At the beginning of the mission they are sent to different locations to minimise power consumption and travelling time.
2.3. Dual System: ASV-Buoy {#sec2dot3-sensors-18-03497}
--------------------------
Taking into account the aforementioned problems, two options are therefore available for long-term missions: (i) deploying a large number of fixed marine observatories (buoys) over the area to be monitored (unfeasible due to the impact on maritime traffic, the environment, tourism and cost); or (ii) autonomous systems able to perform permanent monitoring missions with the minimum cost in temporary deployment of vehicles and teams and maximum observation time.
This article describes the solution we have devised for permanent monitoring of lakes or shallow coastal areas, such as the Mar Menor in southeastern Spain (for a description of which see [Section 4.1](#sec4dot1-sensors-18-03497){ref-type="sec"}). The system combines the main advantages of autonomous systems and buoys to overcome the above-described limitations and challenges. It can thus be described as a "dual solution robot-buoy": an autonomous solar-powered marine robotic observatory equipped for data collection during missions, navigation to inspection points, etc. that also has the ability to anchor and become a "buoy", for either taking measurements at a specific point or prioritizing energy recharge as required, thus avoiding being left to drift with the consequences that it entails (especially near the coast) and returning to monitoring work when the required energy levels are reached. This novel dual robot-buoy configuration ensures a permanent presence in the observation area, as described below. As far as we know, no other similar solutions have yet been proposed.
This robot could also use the Predictive Control Model (PCM) developed by the authors in \[[@B21-sensors-18-03497]\] to obtain optimal battery charging and power consumption, using an algorithm that inhibits fluctuations in the power output of the photovoltaic system due to temperature changes, cloud cover, irradiation, etc. The sinusoidal rotational vector method \[[@B22-sensors-18-03497]\] could also be used to estimate the vessel's attitude; this method was installed in a land vehicle to determine the difference between the vehicle's measured and expected attitude.
3. General Description of the Dual Observatory {#sec3-sensors-18-03497}
==============================================
This section briefly describes the fundamental characteristics of the dual robotic observatory BUSCAMOS-RobObs. The following sub-sections will describe in greater detail its energy system, decision-support system, and the hardware and software architecture. The dual system can be defined as a dual autonomous oceanographic observatory, in which *dual* refers to its being either an autonomous vehicle or an oceanographic observatory, as required.
BUSCAMOS-RobObs is a mono-hull vessel built of polyester reinforced with fiberglass (see [Figure 1](#sensors-18-03497-f001){ref-type="fig"}). The hull is 5.10 m long and 1.97 m wide. The displacement of the vessel under normal operating conditions is approximately 1000 kg, with an average draft of 0.325 m. The vessel is steered by two independent outboard propellers anchored to the transom, where each thruster is powered in series by two 28 V serial connected lithium ion batteries. It is also equipped with a rudder control, which changes the thrust direction of the propellers by means of a linear motor.
The BUSCAMOS-RobObs ASV includes the following sensing devices: a SIMRAD broadband 3G radar, an aft camera (model PTZ AXIS P5534-E), a bow camera (model AXIS P1435-LE), a sidescan sonar (model Tritech SeaKing), an imaging sonar (model Tritech Micron Sonar) and a multi-parametric probe (model YSI 6600 V2). The side scan sonar generates the images of the seabed that are shown in [Section 5](#sec5-sensors-18-03497){ref-type="sec"} (Figure 15), while the imaging sonar and the 3G radar are used for underwater and surface obstacle avoidance. The multi-parametric probe, whose main characteristics are described in \[[@B23-sensors-18-03497]\], can measure, temperature, salinity, density, dissolved oxygen, chlorophyll concentration and nitrates.
The vessel employs a photovoltaic system to produce energy and a set of batteries for the long-term operation of the system. Eight lightweight flexible photovoltaic panels cover an area of 4000 × 1355 mm on the ASV's deck, with a total installed photovoltaic power of 1040 Wp. As can be seen in [Figure 1](#sensors-18-03497-f001){ref-type="fig"}, the panels' position on the hull has little effect on the vessel's aerodynamics and manoeuvrability. [Section 3.2](#sec3dot2-sensors-18-03497){ref-type="sec"} gives further details of the electric system.
The BUSCAMOS-RobObs control hardware architecture is described in [Section 3.3](#sec3dot3-sensors-18-03497){ref-type="sec"}, including a detailed description of the main software architectural layers, their roles and responsibilities, while [Section 3.4](#sec3dot4-sensors-18-03497){ref-type="sec"} describes the ASV's on-board hardware and [Section 3.5](#sec3dot5-sensors-18-03497){ref-type="sec"} the central command, control, communication and computing station. The ASV hardware is composed of three main nodes, interconnected via CAN bus and Ethernet, and with the base station via WiFi, radio and GPRS. The modules are (see Figure 3) as follows:-*Operational Module*, responsible for the navigation, consisting of a National Instrument sbRIO 9606 main controller and two microcontroller-based electronic cards.-*Energy Sensing and Management Module*, consisting of an electronic card based on a micro-controller, which acquires information from the different sensors, checking their status periodically and managing energy. It also controls the anchor deployment and retrieval motor.-*Tactical and Strategic Module*, responsible for operating the decision-making algorithm and storage and management of the acquired data. This module is a single-board DFRobot Lattepanda computer based on the Intel Atom Cherry Trail.
The Energy Sensing and Management Module manages mode changes between autonomous vehicle, fixed oceanographic observatory, or prioritising battery recharging. As described above, one of the dual observatory's special characteristics is that during battery-charging the vessel is anchored to the seabed with minimum power consumption. When the batteries are fully charged the control system wakes up and starts the scanning process until the batteries return to minimum, when it deactivates the systems. The vehicle thus combines the advantages of buoys and ASVs, while it can also adopt the at-rest mode in bad weather or sea conditions.
3.1. Energy Management System {#sec3dot1-sensors-18-03497}
-----------------------------
The photovoltaic system configuration is a typical standalone installation. The ASV uses a photovoltaic generator system with eight inboard flexible photovoltaic solar modules of 130 Wp each. With the propellers in power mode the vessel's systems consume around 1802 Wh. As the photovoltaic solar modules are laid horizontally on the deck, their performance is reduced and they provide around 1040 Wh, insufficient to drive the boat directly from the photovoltaic generator, so that an additional intermediate energy storage system is required. The system stores energy in four batteries 26 V of 104 Amps/unit, distributed in two in-series batteries for each torch propeller, giving 6 h of navigation autonomy, according to Equation (1):$$\text{Autonomy} = \frac{10816\ {Wh}}{1802\ {Wh}} = 6\ h$$
The battery bank provides stable current and voltage by eliminating transients and provides surge currents to the thrusters when required. Two solar charge controllers are placed between the PVs and the battery bank. These devices adjust the charge rates to the status of the battery bank, applying maximum power point tracker (MPPT) technology. This device boosts the PV lower voltage to charge higher voltage lithium batteries up to 54 V nominal. The characteristics of the photovoltaic system are shown in [Table 1](#sensors-18-03497-t001){ref-type="table"}.
3.2. Decision-Support System {#sec3dot2-sensors-18-03497}
----------------------------
The control strategy implemented in the ASV is structured into three software layers, as shown in [Figure 2](#sensors-18-03497-f002){ref-type="fig"}:
*Strategic Level*: is the high-level decision-making subsystem for selecting the best geographical area (the "box") for exploration, according to an analysis of the experimental physical and chemical parameters measured, time elapsed since the previous exploration, energy constraints, and the distance between the present position and the next targeted exploration area. This layer is described in more detail in the following section.
Tactical Level: is the intermediate decision-making system, which receives the next "box" to be explored from the layer above and designs a plan to reach and explore it. It comprises a fuzzy level subsystem to optimize route, speed, course and starting angle. This algorithm processes information from the vessel sensors and the status of the vehicle itself and its environment, such as the tag of the next box to be explored, power available in the batteries, and energy supplied in real time by the on-board photovoltaic system. The GPS coordinates of the next waypoint and thruster speed are sent to the operational level layer, which is the lowest level in the hierarchy.
*Operational Level*: is a low-level purely technical decision-making step focused on navigation control. The system controls the course, speed, angle, etc. and operates the obstacle avoidance sensors. The operational control level uses a fuzzy logic feedback control loop: the data supplied by the tactical level are read and compared with the reference signal, and any errors are used as a control parameter to correct the course.
These decision levels are implemented in a hardware architecture that maintains an equivalent hierarchical order, as described below.
3.3. Control Hardware Architecture {#sec3dot3-sensors-18-03497}
----------------------------------
The control hardware architecture consists mainly of three modules, each with a series of electronic devices and cards for specific functions. All the modules are interconnected via a CANopen fieldbus and an Ethernet network (see [Figure 3](#sensors-18-03497-f003){ref-type="fig"}).
The Operational Module is primarily responsible for vehicle navigation: motor control, GPS acquisition, inertial unit heading acquisition, obstacle detection and safety. It consists of a National Instruments sbRIO 9606 main controller and two specifically designed electronic cards, Nodes 1 and 2 of the CANBUS fieldbus, based on a PIC 18f4685-E/PT microcontroller (8 bits and clock frequency of 25 MHz). Each node electronic system includes modules that guarantee the desired functionality (CAN-TTL transducer, serial communication transducer, precision AD converters, voltage regulators, protections, drivers for motor management, etc.).
The Energy Sensing and Management Module (i) acquires information from the multi-parametric probe and cyclically checks its status (operational, maintenance required, etc.); (ii) controls camera and sonar power and (iii) controls the anchor deployment motor, changes vehicle mode between autonomous vehicle, fixed oceanographic observatory, or prioritization of energy recharging. It consists of two custom-designed electronic cards, one of which is the third node of the CANopen fieldbus, and the other an expansion of the previous one, performing load management functions for the different connected devices. The third node has a hardware structure similar to the other two.
The Tactical and Strategic Module manages the decision-making algorithms and database. Its main element is a single-board DFRobot Lattepanda PC based on the Intel Atom Cherry Trail chip with 4 GB RAM and 64 GB internal memory. This module stores in a MySQL database the readings taken, which are used by the decision-making algorithms and in turn are arranged for synchronization and consultation with land-based communication stations.
Regarding the communications infrastructure, BUSCAMOS-RobObs uses CAN bus, Ethernet network, WiFi and radio link. The CANopen fieldbus communicates all three nodes with the sbRIO master controller, enabling reliable and robust process data exchange between the controller core and the distributed periphery at a sufficiently high speed (500 kbps). The Ethernet network interconnects the sbRIO with the CPU, receiving the instructions and navigation parameters according to the results generated in real time by the decision-making algorithm. The vehicle communicates with the land-based station through WiFi and radio. It is equipped with omnidirectional WiFi antennas and high gain radio. The WiFi network, with its higher bandwidth, is designed to exchange large volumes of data. The low-bandwidth and long-range radio network guarantees communications during missions. For security reasons both networks have redundancy in critical tasks such as receiving instructions from the base, positioning and managing alerts.
3.4. Software Architecture {#sec3dot4-sensors-18-03497}
--------------------------
The software module name is shown next to the name of the main code block algorithm implemented in each platform in [Figure 4](#sensors-18-03497-f004){ref-type="fig"}. The latter label is divided into two parts separated by a hyphen, where the first part indicates the hardware platform on which the software module is running (sbRIO, N1, N2, N3, CPU). The arrows indicate the direction of the data flow, and the label shows the type of communication or connection between the blocks. The layer to which each software element belongs is also displayed.
The hardware elements in the operational layer are basically micro-controllers and the sbRIO. The micro-controllers execute the functions shown in the figure periodically, triggered by time or event. For example, CAN readings every 10 ms, AD conversions every 5 ms, watchdog every 250 ms, RS232 acquisition every 250 ms. The algorithms/functions shown in the figure are the most important ones that each hardware element executes. In the case of software elements running on the CPU (belonging to the Strategic and Tactical layer) on the Windows 10 LTSB operating system, the software modules shown in the figure are programs running on this platform. This platform is powerful enough to run Fuzzy and real-time decision-making algorithms, as well as manage the MySQL database, make periodic communications checks and other lower-priority processes. A brief description of some of the software modules involved in the control is given in [Table 2](#sensors-18-03497-t002){ref-type="table"}.
3.5. Command, Control, Communication and Computation Station: IUNO {#sec3dot5-sensors-18-03497}
------------------------------------------------------------------
The "*Interface for Unmanned Drones*" (IUNO) is the base station control software, on which the human operators plan and send missions to the ASV. It was designed with the purpose of simplifying the management of unmanned vehicles by analysing the data collected by the vehicles and providing a fully automated vehicle management response to guarantee the success of the task. IUNO has a graphic user interface, showing the operator all the navigation information (vehicle location, power consumption and storage, historical tracks and programmed tracks, etc.). A screenshot of the IUNO software IUNO is shown in [Figure 5](#sensors-18-03497-f005){ref-type="fig"}.
IUNO has a wide range of functions, the most important being: multivehicle management, voice synthesizer, alarm event, geographical location and mission planner, fast navigation mode covering the planned geographical location given by coordinates (latitude and longitude), targets, and two navigation modes, manual and semiautomatic. The navigation modes are synchronized with the on-board sonar and GPS localizer and also incorporate OpenCPN nautical charts (.bsb, .kap, .map) to facilitate vehicle management. All the measured parameters are stored and sent to remote databases with a standard Keyhole Markup Language (KML) and Keyhole Markup Zip (KMZ) location marker, widely used for example in Google Earth. Being a command & control station, new missions can be launched by IUNO to any of the vehicles it is coordinating, while at the same time the human operator can cancel any of the missions assigned to any of the vehicles, whether it was previously assigned manually or automatically.
4. Design of Strategic Decision-Making Layer and Application to the Observation Mission in the Mar Menor {#sec4-sensors-18-03497}
========================================================================================================
This section describes the design of the upper strategic decision-making layer, shown in [Figure 2](#sensors-18-03497-f002){ref-type="fig"}. At this level, a fuzzy algorithm processes the information from the vehicle's sensors, its status and the environment to select the next exploration zone. For this, it has an up-to-date grid map of the entire exploration area, with a record of the status of each box, which depends on the time expired since the last exploration and the latest levels of physical and chemical parameters observed or recorded by other means. Among other parameters, the fuzzy system takes into account the vehicle's power status to determine whether it has enough power to reach the next exploration box and when it will arrive, or whether it must anchor to recharge the batteries. To clarify this process, we added to the explanation of the algorithm an example of the exploration area, the Mar Menor, where all the tests and simulations were carried out. We will begin by describing this inland's sea's unique characteristics and the classification of the exploration zones.
4.1. Classification of Exploration Areas in the Mar Menor {#sec4dot1-sensors-18-03497}
---------------------------------------------------------
The Mar Menor, in the southeast of Spain (see [Figure 6](#sensors-18-03497-f006){ref-type="fig"}), is a unique and a valuable resource. Its uniqueness is due to its geographical characteristics (Europe's largest salt lake, with an area of 180 km^2^), with a unique ecosystem in need of conservation. The Mar Menor is a valuable economic resource for the Region of Murcia for the tourism its attracts, for its beaches and good climate in \[[@B24-sensors-18-03497]\] (mean maximum temperature 26 °C, mean minimum temperature 14 °C, 127 days of temperatures above 30°, and 3100 h of sun). This abundant supply of sunlight is an essential resource as an alternative energy source for the robotic mobile observatory.
Due to its unique characteristics, the Mar Menor has attracted the attention of numerous scientific researchers, such as biologists, environmental scientists, oceanographers, etc., especially since 2016, when the rising levels of chemical and biological contamination caused detectable changes in the composition and chemical balance of its waters. Algae proliferated due to the intense levels of human activities, discolouring the waters and affecting the ecosystem, since the clouded waters impeded the penetration of sunlight and this affected its flora and fauna. Tourism was also affected, since its beaches lost all of their previous 19 EU Blue Flags between 2016 and 2018. Blue Flags \[[@B25-sensors-18-03497]\], awarded by the Foundation for Environmental Education (FEE), certify that a beach, marina, or sustainable touristic boat operation meets the Foundation's environmental, educational, safety, and accessibility criteria.
The first step in effectively exploring the Mar Menor was to define a grid map, dividing the entire area into 180 1 km × 1 km boxes, using the National Topographic Map:Hoja 955 Torre-Pacheco, Hoja 956 San Javier and Hoja 978 Llano del Beal 1:50.000 scale (see [Figure 6](#sensors-18-03497-f006){ref-type="fig"}). The boxes were geo-referenced by GPS or UTM geographical coordinates stored in the on-board database. Data from previous exploration missions (historical archive) was also kept, including all the parameters measured by the on-board sensors or other information sources, like buoys or manual data collection. These parameters included environmental, physical, chemical and biological information (nitrate and phosphate contamination), turbidity, acidity, water density, oxygen, suspended particles, etc. All the parameter sets were tagged with their geographical location, date and time of acquisition.
4.2. Classification Strategy of Areas of Interest {#sec4dot2-sensors-18-03497}
-------------------------------------------------
The mission's strategic planning software layer used the information stored on the map to classify the state of each box. To do this, it has a fuzzy algorithm that whose input is the average stored values of the temperature, salinity, turbidity, oxygen, chlorophyll and nitrates variables in each one. On the basis of this information, the boxes were classified with the specific colour of a fuzzy state. To carry out the simulations, the values of environmental, physical, chemical and biological parameters of the 182 boxes were taken from a previous manual data collection in some representative areas and extrapolated to the other boxes, as shown in [Figure 7](#sensors-18-03497-f007){ref-type="fig"}.
The anomalous values of box 31 were manually added for simulation purposes only. The graph shows the values obtained from the 182 boxes, normalised and fitted to a scale from 0 to 1. The values are entered in the Fuzzy Sensor system to obtain the status of each box, as described below. The value for each sensor in each box is calculated as the mean of all the readings in that box (1 km^2^, in total the vessel travels 11 km) to avoid that a single measurement defines its value. [Figure 8](#sensors-18-03497-f008){ref-type="fig"}a depicts the fuzzy membership functions for the input sensor values (in this case, for nitrates values), and for the output value ([Figure 8](#sensors-18-03497-f008){ref-type="fig"}b). This output value defines the state of the grid, and therefore the colour of the area on the map. In total, 64 knowledge rules have been implemented in the fuzzy system, which in a summarized way follow the behaviour shown below:-If the value of any of the inputs is EL or EH, then the output is VeryBad.-If there are no EL or EH values but there is at least one VL or VH value, then the output is Bad.-If there are no EL, EH, VL or VH values but there is at least one L or H value, then the output is Normal.-If there are no EL, EH, VL, VH, L or H values but there is at least one LL or LH value, then the output is Good.-If all input values are N, then the output is VeryGood.
Thus, for a specific experiment, each of the boxes on the map were coloured as shown in [Figure 9](#sensors-18-03497-f009){ref-type="fig"}: Very Good (green), Good (light green), Normal (yellow), Bad (orange) and Very Bad (red). A critical zone in a Very Bad state can be seen in Box 31, at the mouth of the Albujón river, where runoffs of agricultural nitrates reach the sea. Several Bad and Normal areas can also be seen, while the remainder is classified as Very Good.
4.3. Decisions Required to Navigate to the Area of Interest for Buoy-ASV Dual Operations {#sec4dot3-sensors-18-03497}
----------------------------------------------------------------------------------------
The decision on which box to explore is made by taking into account three fundamental parameters: its degree of interest (related to the state of the physical-chemical variables), the distance of the ASV to other squares, and the time expired since the last exploration. Equation (2) gives the rules used for the selection:$$\text{BoxInterest} = \frac{\left( {\text{ObservationInterest} + \text{UpdateInterest}} \right)}{\text{Distance}}$$ where:$$\text{ObservationInterest} = \left\{ \begin{matrix}
{24\left( \text{two\ years} \right)~\text{if\ box\ State\ is\ Very\ Bad}} \\
{12~\left( {{one}\ {year}} \right)~\text{if\ box\ State\ is\ Bad}} \\
{0~\text{otherwise}} \\
\end{matrix} \right.\text{and\ UpdateInterest} = \text{time\ in\ months\ since\ last\ measurement}$$
The system calculates BoxInterest values for the whole map and selects the one with the highest value. Thus, from its initial starting point, the ASV will sail to and explore the indicated area of interest. Should its battery load fall below 20%, the vehicle will stop, drop its anchor and go into stand-by until the solar panels have recharged the batteries to 75%, after which it will wake up and continue the exploration. Should it receive an alert from an area, the State value of that area will become a priority (red) and the vessel will proceed to that area following the procedure outlined above.
During the journey to and exploration of the different zones, the values in the database are constantly updated with information on the 6 physical-chemical parameters and their date of capture.
4.4. Simulations to Validate the Navigation Process and Fuzzy Zone Classification Algorithm {#sec4dot4-sensors-18-03497}
-------------------------------------------------------------------------------------------
This section describes an operational simulation performed on Simulink using the fuzzy classification algorithm and navigation strategy described above. A navigation step from Box 145 to Box 31 is simulated on the Mar Menor grid map, as seen in [Figure 10](#sensors-18-03497-f010){ref-type="fig"}.
The simulation sequence was as follows: while in Box 145 BUSCAMOS-RobObs receives an alert from Box 31, where a high nitrates reading has been detected in the fuzzy map output shown in [Figure 10](#sensors-18-03497-f010){ref-type="fig"}. New *BoxInterest* values are then computed for the boxes ([Figure 11](#sensors-18-03497-f011){ref-type="fig"}b shows only the first 90 boxes). These values are calculated from the three parameters shown in [Figure 11](#sensors-18-03497-f011){ref-type="fig"}a and Equation (2): the first scope shows the values obtained to establish the colour intensities on the map in [Figure 9](#sensors-18-03497-f009){ref-type="fig"}; the second scope shows the values of the distance from the ASV's current position to the other boxes; while the third shows the time that has passed since each box was explored. [Figure 12](#sensors-18-03497-f012){ref-type="fig"} gives the rule set that generates the fuzzy output from the input values. As can be seen, a total of 10 rules were defined to model the box selection procedure.
4.5. Energy Management Simulations {#sec4dot5-sensors-18-03497}
----------------------------------
As the ASV must travel by means of the energy captured by its solar modules, we must know the coordinates of the target box, the time of year, hours of solar radiation received, the power of the vessel's photovoltaic generator, the generator's efficiency, and any losses in the system.
The energy management system is described in [Section 3.1](#sec3dot1-sensors-18-03497){ref-type="sec"}. The energy balance between the generating system (maximum 1040 Wp), battery storage and energy consumption is shown in [Figure 13](#sensors-18-03497-f013){ref-type="fig"} and [Figure 14](#sensors-18-03497-f014){ref-type="fig"}. The photovoltaic energy generated in any month is clearly not enough to cover the power demand, since there are losses in the wiring and electrical systems and batteries, due to temperature changes and the efficiency of the electronic equipment, horizontal panels, etc. The simulation results after considering all these factors are shown in [Figure 13](#sensors-18-03497-f013){ref-type="fig"}, which shows that the power required to supply the system all year round would be 623.8 kWh. Since the system had to be completely autonomous and we only had the battery storage to fall back on, the vehicle was designed as a dual system.
The environmental data (irradiation, daily average temperature) are provided by METEONORM V6.1.0.23, which is a complete database; the data were obtained for San Javier, Murcia, Spain (lat. 37.8 N, long. 0.8 W). The results of the simulations are shown in [Table 3](#sensors-18-03497-t003){ref-type="table"} and [Table 4](#sensors-18-03497-t004){ref-type="table"}.
Since the fully charged batteries can operate for 6 h, the daily exploration of each box was divided into time bands (see [Table 4](#sensors-18-03497-t004){ref-type="table"}), one in the morning, another at noon and the last one in the late afternoon. When recharging batteries and at night, the vehicle remains anchored to the seabed, in stand-by mode.
This sampling strategy was operated through different levels of interconnected fuzzy control techniques to achieve a robust strategic control system and effectively increase the autonomy of the proposed robotic observatory.
5. Experimental Tests and Results {#sec5-sensors-18-03497}
=================================
This section describes in detail a 10-day mission to test the vessel's power system, autonomy and long-term operations. The test began on 20 July 2018 at 8:00 in the morning. Prior to starting the experiment, the different systems were checked as follows: battery power, photovoltaic recharge, manual test of propellers, rudder and anchor deployment, control system auto-diagnosis module by means of the IUNO remote monitoring system, verification of sensors and devices (GPS, multiparameter probe, scanning sonar, etc.), and communications between the vessel and base station.
The vessel was launched with fully charged batteries in Box 102, near the *Los Nietos* Yacht Club (see [Figure 6](#sensors-18-03497-f006){ref-type="fig"}). The database for the decision making algorithms was loaded with previously acquired data from the exploration boxes. The status colours of the boxes and the mission results can be seen in [Figure 16](#sensors-18-03497-f016){ref-type="fig"}. The first box inspected was N° 102, (UTM coordinates: X: 696,000, Y: 4,170,000). The exploration route followed parallel lines separated by 100 m over the entire length of the grid. During the tour, sonar images, videos of the seabed and water parameters were recorded (see [Figure 15](#sensors-18-03497-f015){ref-type="fig"}). The inspection of Box 102 ended at 11:00 a.m., with a power consumption of 2.5 kWh. The values obtained by the multiparameter probe are shown in [Table 5](#sensors-18-03497-t005){ref-type="table"}. These data were stored in the database together with the corresponding timestamp, which can influence the decision on the next box to visit.
Having only updated data of a few nearby areas (green boxes around 102 in [Figure 16](#sensors-18-03497-f016){ref-type="fig"}), the criterion of greatest weight in the decision making was the proximity of contiguous zones. We thus proceeded to inspect in the same way and with the same type of sampling Boxes 123 (UTM coordinates: X: 697,000, Y: 4,170,000) and 145 (UTM coordinates: X: 698,000; Y: 4,170,000), see red flags in [Figure 16](#sensors-18-03497-f016){ref-type="fig"}). Box 123 was completed at 16:00 h, with an energy consumption of 3.8 kWh. The increased power consumption was mostly due to changing course and a strong wind.
Having completed 52% of the inspection of Box 145 at 19:00, a nitrates alert in Box 31 was simulated from the IUNO system in the base station. This susceptible area is at the mouth of a river and should in fact be permanently monitored by a fixed buoy, as it is a known critical dumping area. Before this event, the vessel was on its way to Box 31, as the decision-making algorithms had selected this for monitoring next. After travelling 4 km and consuming 4.7 kWh, the vessel stopped as the battery charge was below 20%. The algorithms decided that the best strategy was to switch to recharge-only mode and anchor overnight at 1.85 km from Box 31.
The vessel remained anchored to recharge above 75% (8917 Wh) until 3:00 p.m. on 23 July and then continued to Box 31, using 1.0 kWh. The exploration of Box 31 began at 16.00 h and ended at 20:00 h, after which it began to explore Box 32, where at 9:00 p.m. it dropped anchor for an overnight stay (see [Figure 16](#sensors-18-03497-f016){ref-type="fig"}).
At 7:00 a.m. on July 24, the exploration of Box 32 resumed but stopped again at 11:00 a.m., when another simulated alert was received from Box 5 (UTM coordinates: X: 690,000; Y: 4,176,000). The ASV the travelled 1 km to that box until the batteries reached 20% and entered recharge mode at UTM coordinates X: 691,284; Y: 4,174,708, where it stayed until 26 July, when the charge rose above 75%. It then went 1.83 km to Box 5, and began an inspection but dropped anchor at 21:00 h. At 7:00 a.m. on 27 July, the ASV resumed the exploration of Box 5, which ended at 8:00 a.m., when it received another alert from Box 69, proceeding to that box. After 2 km the battery level again fell below 20%, when it again entered recharge mode at UTM coordinates: X: 692,000; Y: 4,177,000. It remained here until two days later, 29 July, with the batteries over 75% and went to Box 69, beginning the exploration at 16:00 h, ending at 9:00 p.m. and going into standby. At 7:00 a.m. on 30 July it began exploring Box 89 until the end of the test at 12:00 h, having been working autonomously for 10 days. In [Figure 16](#sensors-18-03497-f016){ref-type="fig"} shows the course followed. The target boxes are marked with red flags.
The final result of the energy balance can be seen in the graph in [Figure 17](#sensors-18-03497-f017){ref-type="fig"}, which in addition to the ASV consumption (orange curve, in W) shows the power generated by the photovoltaic system (blue curve, in W), the energy stored in the batteries (green curve, in Wh), the times when the ASV sets sail, and when the anchor is dropped to recharge overnight. Note that though the units are different, the scale is the same for the curves. The white anchor symbol represents a stop to recharge batteries, a black anchor means an overnight stop, and the propeller shows the time when the boat gets under way. The horizontal dashed lines represent the minimum battery charge to anchor the vessel (in red, 20%), and the minimum charge to get under way (in green, 75%).
6. Conclusions and Future Works {#sec6-sensors-18-03497}
===============================
This article has described the novel BUSCAMOS-RobObs dual robot-buoy mobile observatory, which was specifically designed to maintain a permanent presence in shallow waters, by combining the main advantages of autonomous vehicles and buoys to overcome their limitations and challenges. Although many solutions have been proposed in the literature for monitoring marine environments, as far as we know none of them satisfies the power requirements for long-term explorations.
Since an autonomous power system is fundamental in achieving a permanent presence, the article gives an in-depth analysis of the system's unique power management system, including its design (1040 Wp photovoltaic generator, 8 solar panels and 10,740 Wh batteries), a simulation of its seasonal performance (summer battery recharging is 50% faster than winter). Actual data on BUSCAMOS-RobObs energy consumption and energy management was used in an experimental 10-day mission in July 2018 in Europe's largest inland sea, the Mar Menor, with 186 Km^2^ in the southeast of Spain.
BUSCAMOS-RobObs is a dual system because it combines the advantages of ASV mobility with the stability of a fixed buoy. As an autonomous vehicle it can travel to selected points to measure water quality and scan the seabed. When acting as a buoy it can anchor over a point either to take continuous measurements or recharge its batteries. The ability to recharge batteries when anchored to the seabed not only means faster recharging, since the position control systems are on standby, but also ensures that the vehicle will not drift due to currents or winds. This ability is what gives the system an enormous autonomy to carry out long-term missions.
As we concluded from the review of the state-of-the-art, BUSCAMOS-RobObs provides a novel solution and achieves a permanent autonomous presence in lakes or shallow coastal waters. This improves the autonomy in monitoring water quality parameters and avoids the problems associated with the deployment of a large number of fixed marine observatories (buoys), which influence and impact maritime traffic, the environment, tourism and the cost involved.
Another notable contribution, in line with the objective of minimising the vehicle's energy consumption, is that its fuzzy mission management system takes into account the status of the vehicle's batteries and the distance to alternative exploration areas to select the next mission zone. Also described was the fuzzy system that uses the measurements of the vehicle's sensors to characterise the water quality and possible contamination to generate a colour map of the status of the area, as described in [Section 5](#sec5-sensors-18-03497){ref-type="sec"}. This map is updated as new measures are taken, and is also considered when determining the next area to explore.
The viability of the approach was demonstrated by a real 10-day monitoring mission carried out in the Mar Menor, and by several simulations of the vehicle's performance and power consumption. The results of both the simulations and the experimental mission were found to be satisfactory, reaching the objectives of maintaining a permanent monitoring presence for a period of 10 days and travelling 92.28 Kms without human intervention. The data collected were monitored and stored remotely in the IUNO system. The energy management system also performed well, being able to navigate for the entire period, anchoring each night and selecting the 4 stops for recharging (see [Figure 17](#sensors-18-03497-f017){ref-type="fig"}).
No limitations were observed in the operation, apart from the one derived from the scope of BUSCAMOS-RobObs application. Since the anchoring depth limit is 8 m, it cannot operate in deeper waters, which is why we specify that it is a solution for permanent monitoring of shallow waters. It is also necessary to take into account the prevailing weather conditions (especially wind), wave heights and tides in the area, which may hamper operations even with recharged batteries. Still, it is a solution that can be applied to permanently monitor a multitude of lagoons and coastal seas around the world.
Given the large area to be covered in monitoring the Mar Menor (186 Km^2^) and the speed of the vessel, we estimate that it would take up to 1 year to complete a full autonomous exploration. In the future, deploying more vehicles could reduce this time. The number of vehicles depends on the maximum allowed age for the data gathered and the available budget. For example, we estimate that 3 vehicles could cover the whole area in 3--4 months. However, it should be noted that this is a monitoring vehicle and is not equipped to intervene, and therefore it is acceptable that it takes a while for the vessel to reach a box in which an alarm has raised. Should a spill happen, other means must be deployed to deal with it.
Regarding future work, we are currently planning to: (i) perform longer monitoring missions to improve the efficiency of the solar capture of the photovoltaic system and improve the capacity of the batteries; (ii) carry out missions in autumn and spring with fewer hours of daylight and more cloud cover in less favourable areas; (iii) manage recharge periods by means of new algorithms so as to start the vehicle's activity with lower power levels; (iv) introduce new sources of renewable energy, such as wind turbines, and (v) automate post-processing of data using big data techniques.
The authors would like to thank the Spanish Navy for ceding a Pluto Unmanned Underwater Vehicle and by projects TIN2014-55024-P from the Spanish Ministry of Science and Innovation and P11-TIC-8001 from Junta de Andalucía (including FEDER funds), "Research Programme for Groups of Scientific Excellence at Region of Murcia" of the Seneca Foundation (Agency for Science and Technology of the Region of Murcia---19895/GERM/15).
I.G.-R. has done this research as part of his Ph.D. work under the supervision of F.J.O. and A.G.-G. J.C.M.-M. has done this research as part of his Ph.D. work under the supervision of A.G.-G. F.J.O. and D.A. supervised the development of software architecture of the system. All authors were involved in discussions over the past year that shaped the article in its final version.
This work was partially supported by the BUSCAMOS Project under the program DN8644 COINCIDENTE of the Spanish Defense Ministry.
The authors declare no conflict of interest.
{#sensors-18-03497-f001}
{#sensors-18-03497-f002}
{#sensors-18-03497-f003}
{#sensors-18-03497-f004}
{#sensors-18-03497-f005}
{#sensors-18-03497-f006}
{#sensors-18-03497-f007}
{#sensors-18-03497-f008}
{#sensors-18-03497-f009}
{#sensors-18-03497-f010}
{#sensors-18-03497-f011}
{#sensors-18-03497-f012}
{#sensors-18-03497-f013}
{#sensors-18-03497-f014}
{#sensors-18-03497-f015}
{#sensors-18-03497-f016}
{#sensors-18-03497-f017}
sensors-18-03497-t001_Table 1
######
Parameters describing the PV modules and batteries.
PV Modules: Enecom Italia, HF130 Batteries: Lithium-Manganese, Torqeedo, Power 26--104
---------------------------------- ------------------------------------------------------- ------------------------- -----------------------------------
Solar power 130 (Wp) Voltage 25.9 (V)
V~oc~ 20.01 (V) Nominal load 104 (Ah)
V~mpp~ 16.86 (V) Total units 4 batteries (2 series/2 parallel)
I~sc~ 8.61 (A) **Total energy stored** **10,740 kWh**
I~mpp~ 8.10 (A)
Total units 8 modules (4 series/2 parallel)
**Total power** **1040 Wp**
sensors-18-03497-t002_Table 2
######
Description of software modules.
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\-**N1_GPSUI**: obtains and validates GPS position and course of the inertial unit-**N2_ObstacleDetector**: detects obstacles and sends alerts to sbRIO-**N3_LoadsEnergyMngr**: manages expendable loads and energy modes (buoy, recharge, ASV)-**N3_SensorAdq**: acquires and supervises the multi-parametric probe and sonars-**N3_AnchorMngr**: manages the anchor depending on the active energetic mode \-**CPU_DB**: manages the connection with the MySQL database and the data stored on it (geo-referenced navigation and sensor data), which is then used by the decision-making algorithms-**CPU_CommSupervisor**: manages communication link with the base station-**CPU_CameraMngr**: manages the on-board cameras
\-**sbRIO_IUNOFuzzyMainController**: set of software implementing the previously described operational level through the following functions: ○**sbRIO_FuzzyHeadingPropellers**: implements fuzzy algorithms to correct the ASV's heading and speed, as well as an auto-adaptable K factor for PIDs**sbRIO_MissionTracker**: supervises the route of the mission. Together with the former, they both constitute the fuzzy navigation controller**sbRIO_ASVEnergyMngr**: energy management controller**sbRIO_ObstEvas**: generates a new path when an obstacle is detected by node N2.**sbRIO_IncidentMngr**: sends incidents to the base station (errors and alerts), depending on their severity. Manages self-restoration of some solvable incidents**sbRIO_CmdMngr**: manages the connection and interface with the base station, sending and receiving commands
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
sensors-18-03497-t003_Table 3
######
Summary of the results of the simulation obtained with PVSYS.
Annual Recharges Energy for Recharging Annual Power Requirements Useful Solar Power Energy Gap Performance Ratio (PR)
------------------ ----------------------- --------------------------- -------------------- ------------ ------------------------
143 10.816 kWh 1558 kWh 934 kWh/year 624 kWh 51.8%
sensors-18-03497-t004_Table 4
######
Box exploration time bands, stoppages and recharging periods in different seasons of the year.
-----------------------------------------------------------------------------------
Season Daily Exploration Periods Recharging Periods Night Anchorage Periods
-------- --------------------------- -------------------- -------------------------
Summer 07:00--08:00\ 08:00--12:00\ 21:00--07:00
12:00--14:00\ 14:00--18:00\
18:00--20:00 20:00--21:00
Spring 08:00--09:00\ 09:00--12:00\ 19:00--08:00
12:00--14:00 14:00--19:00
Autumn 07:30--09:30\ 09:30--12:00\ 19:30--07:30
12:00--14:00 14:00--19:30
Winter 08:30--09:30\ 09:30--12:00\ 17:00--08:30
12:00--14:00 14:00--17:00
-----------------------------------------------------------------------------------
sensors-18-03497-t005_Table 5
######
Average parameter values in Box 102.
Data Average Value
------------- ------------------
Temperature 21.75 °C
Salinity 42.25 P.S.U
Density 0.12 mg/L
Oxygen 6.84 mg/L
Chlorophyll 2.28 µg/L
Nitrates 0.98 mg N-NO3-/L
| {
"pile_set_name": "PubMed Central"
} |
2.. Introduction {#s1}
================
This article describes a simple formula for calculating the mutual information between random variables where one or both of the variables take values in a metric space. This is relevant to neuroscience because electrophysiological data, whether spike trains from single neurons or collections of spike trains from a population of neurons, can be naturally considered to take values in a metric space \[[@RSOS140391C1],[@RSOS140391C2]\]. It is, in turn, useful to be able to calculate information theory quantities for these data as part of an investigation into effective coding theories of neurodynamics or as a tool for quantifying the relationship between the activity of different neurons or different neuronal populations. However, the relevance is not limited to spike trains, it extends to other electrophysiological data types, such as calcium or electroencephalogram traces. Indeed, non-coordinate metrics or similarity measures are also used, for example, in genetics and biochemistry \[[@RSOS140391C3],[@RSOS140391C4]\], in image analysis \[[@RSOS140391C5]\] and in information retrieval \[[@RSOS140391C6]\].
The aim here is to address two difficulties associated with estimating mutual information on metric spaces. The first difficulty is that information theory is most typically applied to problems where the data are either discrete or take values in an integrable manifold. The second difficulty is that many approaches to estimating information theory quantities can demand an unrealistically large amount of data. The second of these difficulties has been addressed in the past by the Kozachenko--Leonenko estimator \[[@RSOS140391C7]--[@RSOS140391C10]\], but, in line with the first difficulty, this estimator is derived specifically for an integrable manifold or using a local effective dimension. The aim here is to provide a simple approach to a Kozachenko--Leonenko estimator which applies to metric spaces. With this extension, mutual information can be calculated for the broad class of important data where there is a similarity measure, but no coordinates.
The particular application which motivates this article is the problem of calculating information theory quantities for spike trains. Three different types of neuroscience experiment could be considered typical. In the first, the activity of a neuron or group of neurons is recorded from the brain of an anesthetized or restrained animal while that animal is being presented with a series of stimuli. The challenge is to estimate the mutual information between the stimulus and the neuronal activity during the presentation. In the second, neuronal activity is recorded while an animal is moving freely in an arena and the mutual information is to be estimated between the position of the animal at a given time and the neuronal activity in a temporal window centred on that time. In the third example, the mutual information is to be estimated between temporal slices of the spike trains produced by different neurons so that this can be used to measure the relationship between those neurons, for example, at different times during development.
Estimating mutual information is not straightforward in any of these examples because there is no obvious coordinate system for describing neuronal activity. One approach to solving this difficulty is to discretize the spike trains, turning individual fragments of spike train into sequences of ones and zeros with each bit accounting for the presence or absence of a spike in a corresponding time slot \[[@RSOS140391C11]\]. However, the amount of possible words is huge and so this approach is bedevilled by the large amount of data it requires. One response to this problem is to exploit what is sometimes called the birthday problem and to look at coincidences \[[@RSOS140391C12],[@RSOS140391C13]\]. However, the proximity structure of the space of spike trains is poorly approximated by examining the coincidence whereby two spike trains are discretized to the same sequence of ones and zeros. Here, it is proposed that, instead, one of the many metrics or similarity measures on the space of spike trains be used to define proximity \[[@RSOS140391C14]--[@RSOS140391C20]\].
3.. Method {#s2}
==========
Two simple formulae for mutual information are presented in this article, one for the mutual information between a discrete space and a metric space and one for the mutual information between two metric spaces. These two formulae are intended to cover, respectively, the first and the second and third neuroscientific examples described above. There are two main steps to deriving these formulae. Firstly, probabilities are estimated using a simplified version of the Kozachenko--Leonenko approach \[[@RSOS140391C7]--[@RSOS140391C10]\]; this estimate of probability involves terms that depend on the volumes. In the second step, these volumes are estimated using the probability distribution as a measure.
3.1. A formula for the entropy {#s2a}
------------------------------
Consider estimating entropy for a random variable *X* which takes values in a space $\mathcal{X}$ with probability mass density *p*~*X*~(*x*). Given a set of *N* outcomes, {*x*~1~,*x*~2~,...,*x*~*N*~}, the entropy is estimated by $$H(X) \approx - \frac{1}{N}\sum\limits_{i = 1}^{N}\log_{2}p_{X}(x_{i}).$$The problem here is how to calculate this quantity when *p*~*X*~(*x*) is not known. To do this, the approach given in \[[@RSOS140391C7],[@RSOS140391C9]\] is followed in spirit but modified to avoid any quantities that rely on coordinates; the aim is to derive a formula for metric spaces. This will require that a region *B*(*x*~*i*~,*V*) with volume *V* is chosen around each data point; this region will ultimately be specified using the metric, but for now it is supposed only that there is a such a region for each data point.
Now, consider the probability *P*~*k*~(*x*~*i*~) that the region *B*(*x*~*i*~,*V*) contains precisely *k* points; it is $$P_{k}(r_{i}) = \begin{pmatrix}
N \\
k \\
\end{pmatrix}F_{i}^{k}{(1 - F_{i})}^{N - k},$$where *F*~*i*~ is the probability mass contained in *B*(*x*~*i*~,*V*). This means that $$\langle k\rangle = NF_{i}.$$This quantity can be estimated from the data $$\langle k\rangle \approx \#\lbrack B(x_{i},V)\rbrack,$$where \#\[*B*(*x*~*i*~,*V*)\] denotes the number of data points in *B*(*x*~*i*~,*V*); that is for any $B \subseteq \mathcal{X}$ $$\#\lbrack B\rbrack = |B \cap \{ x_{1},x_{2},\ldots,x_{N}\}|.$$This means that $$NF_{i} \approx \#\lbrack B(x_{i},V)\rbrack.$$Now, the probability mass function is approximated by assuming that it is constant in the ball $$F_{i} \approx Vp_{X}(x_{i}).$$This means $$NVp_{X}(x_{i}) \approx \#\lbrack B(x_{i},V)\rbrack$$so $$\log_{2}p_{X}(x_{i}) \approx \log_{2}\#\lbrack B(x_{i},V)\rbrack - \log_{2}N - \log_{2}V$$or $$H(X) \approx \log_{2}N + \log_{2}V - \frac{1}{N}\sum\limits_{i = 1}^{N}\log_{2}{\#\lbrack B(x_{i},V)\rbrack}.$$This assumption is the same as the one used in \[[@RSOS140391C7],[@RSOS140391C9]\]; in \[[@RSOS140391C9]\], some care is given to justifying this as an approximation; here, though, it is introduced only with the general justification that the variation in *p*~*X*~(*x*) should be modest if the region spanned by *B*(*x*~*i*~,*V*) is small.
The formula for the entropy, equation ([3.10](#RSOS140391M3x10){ref-type="disp-formula"}), is similar in spirit to the one given in \[[@RSOS140391C7],[@RSOS140391C9]\]. However, it is not identical, and is, in fact, simpler, because here the probability is estimated using the expected number of points in a ball rather than the size of the ball that contains a given number of points; the latter requires the trinomial, as opposed to the binomial, expansion. Also, of course, the approach here is chosen because it makes it possible to avoid quantities that are only defined on integrable manifolds.
3.2. Estimating the volume {#s2b}
--------------------------
The problem with this is that there may not be an obvious measure. Certainly in the case of spike trains there are no good coordinates and so there is no way to calculate the volume of a region based on the usual sort of coordinate-based measure. However, a probability distribution always defines a measure: the volume of a region can be defined as being equal to the probability mass it contains, $$\text{vol}\, B = P(x \in B).$$This volume can be estimated from the data $$\text{vol}\, B \approx \frac{\#\lbrack B\rbrack}{N}.$$
This gives a trivial estimate of the entropy: differential entropy is zero if the probability distribution is used as the measure and the approximation used here is exact in this case: if *V* =*h*/*N* for some integer *h*≤*N* $$H(X) \approx \log_{2}N + \log_{2}\frac{h}{N} - \frac{1}{N}\sum\limits_{i = 1}^{N}\log_{2}h = 0$$as \#\[*B*(*x*~*i*~,*h*/*N*)\]=*h* by definition. Thus, using the probability as a measure for estimating information theory quantities would be useless if the aim was to estimate entropy. In fact, the entropy is a measure-dependent quantity whose value changes if the measure is changed. This is perhaps less important when a particular relevant coordinate system distinguishes a measure, but generally the significance of the entropy is not clear. The mutual information, however, does not suffer from this problem, its value is independent of the measure used. Furthermore, from the perspective of the approach taken in this article, it involves more than one distribution which means that one distribution can be used to give a measure while calculating estimates for the other distributions.
Here, two cases will be considered: in the first case one random variable is discrete and the other takes its values in a metric space; in the second case both random variables take values in metric spaces. In addition, an estimate is derived for the Kullback--Leibler (KL) divergence between two distributions over the same metric space.
3.3. The mutual information where one random variable is discrete {#s2c}
-----------------------------------------------------------------
In many electrophysiological experiments, stimuli from a discrete corpus are presented repeatedly while spike trains are recorded. In this situation, the stimuli are represented by a discrete random variable and the response by a random variable taking values in a metric space. This situation is considered here; the situation is more general than the neuroscientific application, but, for convenience, the terminology is based on that application.
Let $\mathcal{S}$ be a discrete set representing the stimuli and let $\mathcal{R}$ be a set of responses, which may be spike trains or sets of spike trains recorded from multiple neurons. For simplicity, consider the situation where each element of $\mathcal{S}$ is presented an equal number of times, *n*~t~. Let $n_{s} = |\mathcal{S}|$ be the number of stimuli, so the total number of data points is *N*=*n*~t~*n*~S~.
The metric on $\mathcal{R}$ is used to define the regions that are required around the data points. For a point *r* in $\mathcal{R}$ define an open ball $$B_{\epsilon}(r) = \{ t \in \mathcal{R}:d(r,t) < \epsilon\}.$$Next, let *B*(*r*,*V*) be the open ball *B*~*ϵ*~(*r*) with *ϵ* chosen so that *B*~*ϵ*~(*r*) has volume *V* . The total probability *p*~*R*~(*r*) will be used as a measure and the volume *V* fixed at *V* =*h*/*N* for some *h*≤*N*. This means that *B*(*r*,*h*/*N*) is the open ball around *r* which contains *h* points.
With this measure *H*(*R*)=0. The same measure can be used for the conditioned probabilities; that is, for calculating *H*(*R*\|*S*=*s*) using the conditioned probability *p*~*R*\|*S*=*s*~(*r*). Hence, $$\begin{matrix}
{H(R|S = s)} & {\approx \log_{2}n_{t} + \log_{2}\frac{h}{N} - \frac{1}{n_{t}}\sum\limits_{i = 1}^{N}\log_{2}{\#\left\lbrack {B\left( {r_{i},\frac{h}{N}} \right)} \right\rbrack}} \\
& {\approx - \log_{2}n_{s} + \log_{2}h - \frac{1}{n_{t}}\sum\limits_{i = 1}^{N}\log_{2}{\#\left\lbrack {B\left( {r_{i},\frac{h}{N}} \right)} \right\rbrack.}} \\
\end{matrix}$$Thus, averaging over $s \in \mathcal{S}$ $$\begin{matrix}
{I(R;S)} & {\approx \log_{2}n_{s} - \log_{2}h + \frac{1}{N}\sum\limits_{i = 1}^{N}\log_{2}{\#\left\lbrack {B\left( {r_{i},\frac{h}{N}} \right)} \right\rbrack}} \\
& {\approx \frac{1}{N}\sum\limits_{i = 1}^{N}\log_{2}{\frac{n_{s}\#\lbrack B(r_{i},h/N)\rbrack}{h}.}} \\
\end{matrix}$$
This formula is the same as the one proposed in \[[@RSOS140391C21]\]. However, it is derived there in a convoluted way which leaned heavily on intuition, whereas here the derivation is straightforward and can be easily extended to the case where both $\mathcal{S}$ and $\mathcal{R}$ are metric spaces. It is pointed out there that although the estimate given in \[[@RSOS140391C9]\] is derived using coordinate-based quantities, it can be used to give a formula that applies in this case. Numerical experiments in \[[@RSOS140391C21]\] comparing the two formulae gave very similar results.
3.4. The mutual information where both random variables take values in metric spaces {#s2d}
------------------------------------------------------------------------------------
If $\mathcal{S}$ and $\mathcal{R}$ are both metric spaces, the marginal probability mass functions *p*~*R*~(*r*) and *p*~*S*~(*s*) give volume measures and with these measures *H*(*R*)=*H*(*S*)=0. These same measures also induce a measure on $\mathcal{R} \times \mathcal{S}$, the space where (*R*,*S*) takes its values. In other words, *p*~*R*,*S*~(*r*~*i*~,*s*~*i*~) is estimated by considering regions around (*r*~*i*~,*s*~*i*~) whose volumes are calculated using the measure *p*~*R*~(*r*)*p*~*S*~(*s*) induced from the marginal spaces $\mathcal{R}$ and $\mathcal{S}$. Thus, a square is used to define the regions: $$S\left( {r_{i},s_{i},\frac{h_{1}}{N},\frac{h_{2}}{N}} \right) = \left\{ {(r,s) \in \mathcal{R} \times \mathcal{S}:r \in B_{R}\left( {r_{i},\frac{h_{1}}{N}} \right),\, s \in B_{S}\left( {s_{i},\frac{h_{2}}{N}} \right)} \right\},$$where *h*~1~/*N* and *h*~2~/*N* are the volumes chosen for $\mathcal{R}$ and $\mathcal{S}$. Now, under the induced measure $$\text{vol}\, S\left( {r_{i},s_{i},\frac{h_{1}}{N},\frac{h_{2}}{N}} \right) = \text{vol}\, B_{R}\left( {r_{i},\frac{h_{1}}{N}} \right)\text{vol}\, B_{S}\left( {s_{i},\frac{h_{2}}{N}} \right) \approx \frac{h_{1}h_{2}}{N^{2}}$$so $$I(R;S) \approx \frac{1}{N}\sum\limits_{i = 1}^{N}\log_{2}\frac{N\#\lbrack S(s_{i},r_{i},h_{1}/N,h_{2}/N)\rbrack}{h_{1}h_{2}}.$$Thus the mutual entropy depends on \#\[*S*(*s*~*i*~,*r*~*i*~,*h*~1~/*N*,*h*~2~/*N*)\] which counts the number of stimulus--response pairs (*s*,*r*), where *s* is one of the *h*~1~ points closest to *s*~*i*~ and *r* is one of the *h*~2~ points closest to *r*~*i*~.
It is instructive to consider what happens when the two variables are independent. By definition, *B*~*R*~(*r*~*i*~,*h*~1~/*N*) contains *h*~1~ points out of *N*; as *R* and *S* are independent this means the average number of points in *B*~*S*~(*s*~*i*~,*h*~2~/*N*) which are also in *B*~*R*~(*r*~*i*~,*h*~1~/*N*) is *h*~1~*h*~2~/*N* so $$\frac{N\#\lbrack S(s_{i},r_{i},h_{1}/N,h_{2}/N)\rbrack}{h_{1}h_{2}} \approx 1.$$
The formula includes two integer parameters, *h*~1~ and *h*~2~, which need to be chosen; large values for these parameters reduce the accuracy of the approximation in equation ([3.7](#RSOS140391M3x7){ref-type="disp-formula"}) where the probability is taken as constant throughout the region, whereas taking a smaller value reduces the accuracy of the approximation in equations ([3.6](#RSOS140391M3x6){ref-type="disp-formula"}) and ([3.12](#RSOS140391M3x12){ref-type="disp-formula"}) where the mean or volume is estimated by counting.
3.5. The Kullback--Leibler divergence on a metric space {#s2e}
-------------------------------------------------------
The approach described in this article also gives an estimate for the KL divergence. Consider two random variables *R* and *S* on a metric space $\mathcal{X}$ with probability mass functions *p*~*R*~(*x*) and *p*~*S*~(*x*). If {*r*~1~,*r*~2~,...,*r*~*M*~} and {*s*~1~,*s*~2~,...,*s*~*N*~} are sampled from *R* and *S* then the KL divergence is estimated by $$d(R|S) \approx \frac{1}{M}\sum\limits_{i = 1}^{M}\log_{2}\frac{p_{R}(r_{i})}{p_{S}(r_{i})}.$$Now, as before $$MVp_{R}(r_{i}) \approx \#\lbrack B(r_{i},V)\rbrack.$$However, in this case, the other distribution on the same space is used to measure the volume. If the volume is chosen as *h*/*N*, then *B*(*r*~*i*~,*h*/*N*) is the ball around *r*~*i*~ chosen to be large enough to include *h* points from {*s*~1~,*s*~2~,...,*s*~*N*~} and \#\[*B*(*r*~*i*~,*h*/*N*)\] is the number of points from {*r*~1~,*r*~2~,...,*r*~*M*~} in the ball. The usual formula then gives $$d(R|S) \approx \frac{1}{M}\sum\limits_{i = 1}^{M}\log_{2}\frac{N\#\lbrack B(r_{i},h/N)\rbrack}{Mh}.$$It is easy to check that this formula gives an alternative derivation of the formula above for the mutual information between two random variables on metric spaces.
4.. Conclusion {#s3}
==============
The formula derived here is very simple and is derived without any reference to coordinates on the sample space. It is intended that this demonstrates that the Kozachenko--Leonenko approach applies to metric spaces. The Kozachenko--Leonenko formula presented in \[[@RSOS140391C9]\] relies on a manifold structure in its derivation, but in its final form it is also applicable to metric spaces. It seems unlikely that the performance of the estimates here will be different from the performance of that formula.
In the estimates here, the distance or similarity measure is only required to order the points. Although electrophysiological data are used as a paradigmatic example, the formula can be applied to any pair of random variables taking values in metric spaces or indeed any space with a similarity or distance function suitable for defining regions surrounding each data point. As such it could be used in a straightforward way to calculate, for example, the mutual information between local field potentials and spike trains, or between the position of an animal in a maze and a neuronal population response, or, indeed, between two collections of images, or between images and text.
The estimate provided here relies on an approximation in which probabilities are replaced by counting how many data points fall within a region in space. This makes sound intuitive sense, but it has not been proved here that the estimate is a good one. Indeed, no principle has been described that would allow the volume of the regions to be chosen to sensibly balance the two competing requirements: small regions to reduce the error in assuming the probability is constant throughout the region and large regions so that counting-based estimates are robust. There are two ways in which these difficulties will need to be addressed: theoretically, in demonstrating that the estimate converges under sensible conditions, and practically, in demonstrating that the estimate is accurate and not overly sensitive to the choice of volumes for the sorts of data that are likely to be of interest. Hopefully, the simplicity of the approach described here will aid further development.
I am grateful to Karoline Weisner and Alonso Espinosa Mireles De Villafranca for useful discussion, and to Jonathon Victor, who suggested the extension to the Kullback--Leibler divergence.
Funding statement {#s4}
=================
I am grateful to the James S McDonnell Foundation for financial support through a Scholar Award in Cognitive Science and to the Elisabeth Blackwell Institute for a Senior Fellowship.
Conflict of interests {#s5}
=====================
I declare I do not have any competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Bleeding following cardiac surgery requiring the use of cardiopulmonary bypass (CPB) is associated with prolonged hospital admission and increased mortality \[[@R1],[@R2]\]. This may be a consequence of more complex surgery or because of the increased number of red cell units transfused \[[@R3],[@R4]\]. The latter has been independently associated with an increase in morbidity in a number of studies \[[@R5]--[@R7]\]. Although surgical causes of bleeding will always need to be addressed, bleeding caused by haemostatic failure requires the use of the best available haemostatic treatments with the objective of arresting the bleeding at an early stage, thus reducing the overall number of red cells that are transfused. Two recent studies have suggested that lower thrombin generation measured both pre-CPB and post-CPB is associated with excess postoperative bleeding \[[@R8],[@R9]\]. Therefore, interventions to enhance thrombin generation may be as important as maintaining an adequate platelet count and fibrinogen concentration in controlling and preventing bleeding.
For patients who bleed because of impairment of thrombin generation, the currently available haemostatic options consist of fresh frozen plasma (FFP) and the off-label use of recombinant FVIIa (rFVIIa) and prothrombin complex concentrate (PCC). FFP is the most commonly used and readily available way to replace coagulation factors. However, the volume of FFP needed to produce a clinically relevant increase in clotting factors has been reported to be as much as 30 ml/kg \[[@R10]\]. In the context of surgery involving CPB, this constitutes a considerable volume load in a group of patients who may already have compromised cardiac function. In addition, the transfusion of FFP can be complicated by transfusion-related acute lung injury, and transfusion of larger volumes has been associated with an increased risk of stroke in some studies \[[@R11]\].
rFVIIa has been used off label and in clinical trials in the setting of massive haemorrhage, but evidence to demonstrate efficacy and safety has been limited \[[@R12]--[@R14]\]. Furthermore, the optimal dose is unknown and lower doses than those used to treat haemophilia-related bleeding may be preferable to minimize any risk of thrombosis, especially if these doses were shown to be effective in correcting thrombin generation. PCCs contain FII, FIX and FX or FII, FVII, FIX and FX depending on the manufacturer and the licensing jurisdiction. Successful off-label use of PCCs to manage excess bleeding following cardiac surgery has been reported in a number of studies \[[@R15]--[@R17]\]. However, the optimal dose is also unknown and disseminated intravascular coagulation has been reported when using high concentrations in a porcine trauma model \[[@R18]\], although this may relate to the type of PCC used.
Tissue factor pathway inhibitor (TFPI) is an important regulator of the initiation phase of coagulation. Heparin, the principal anticoagulant used during CPB, induces the release of TFPI from the endothelial surface \[[@R19]\]. In-vitro studies have shown that in the presence of TFPI, the rate of thrombin generation is reduced in a concentration-dependent manner \[[@R20]\]. Recently, a number of studies have reported using TFPI-inhibition to improve thrombin generation in both in-vitro and in-vivo models of haemophilia \[[@R21]--[@R25]\]. This raises the possibility that inhibition of TFPI may be a therapeutic target in treating bleeding in other circumstances, including cardiac surgery.
Thrombin generation assays may offer a better assessment of haemostatic function than routine coagulation tests such as the prothrombin time (PT) and activated partial thromboplastin time (APTT), and thus a better way of measuring the response to any treatment to correct haemostasis. In this study, we investigated the effects of FFP, rFVIIa, PCC and TFPI inhibition on thrombin generation when added *ex vivo* to the plasma of patients who had undergone cardiac surgery requiring CPB.
Methods
=======
Patients were recruited who were undergoing heart valve surgery with or without coronary artery bypass grafting and procedures on the aorta. Informed consent was obtained and the study received approval from the South West Wales local research ethics committee (reference 11/WA/0215). The following demographic data were recorded: age, weight, sex, anticoagulant and antiplatelet medication history, type of operation, duration of aortic cross-clamping, duration of CPB, time and dose of heparin administration, time and dose of protamine given, volume and time of intravenous crystalloid, colloid and blood products administered once preoperative blood samples had been taken. The volume of cell salvage blood was also recorded. The CPB circuits and priming fluids were the same in all cases. Unfractionated heparin was used as an anticoagulant to maintain the activated clotting time more than 400 s. Protamine at a dose of 1 mg per 100 U of heparin was given after the end of the CPB prior to the removal of the arterial and venous cannulae.
Coagulation factor assays and full blood count measurement
----------------------------------------------------------
Whole blood samples were taken into vacutainer tubes containing 3.2% trisodium citrate (Greiner Bio-One, Stonehouse, UK) and ethylenediaminetetraacetic acid (BD, Oxford, UK). Samples were taken before heparin administration and 30 min after reversal of heparin by protamine sulfate, PPP was prepared by centrifuging samples twice at 1650*g* before freezing in aliquots at −80°C for testing later. Full blood cell counts were performed on an ABX Pentra DX 120 automated analyser (Horiba Medical, Northampton, UK). The PT, APTT, Clauss fibrinogen and factors II, V, VII, VIII, IX, X, XI, antithrombin, protein C, free protein S and postoperative anti-Xa activity were measured on an ACL 500 Top (Instrumentation Laboratory, Cheshire, UK) automated coagulometer using standard manufacturer protocols and reagents. Factor XIII activity was measured in a flat-bottomed Immulon 2hb 96-well plate (Diagnostica-Stago, Asnières sur Seine, France) using a chromogenic assay kit from Technoclone (Vienna, Austria) and light absorbance was measured using a plate reader (BioTek, Winoosi, Vermont, USA).
Tissue factor pathway inhibitor ELISA
-------------------------------------
Quantification of full-length and total TFPI was performed using an ELISA technique as described previously \[[@R22]\]. Briefly, full-length and total TFPI were captured by an antic-terminus and anti-KD2 antibody respectively (both from Sanquin Blood Supply, Amsterdam, the Netherlands). A rabbit polyclonal antihuman TFPI antibody was used as the first detection antibody (American Diagnostica, Lexington, Massachusetts, USA). An antirabbit IgG peroxidase conjugate was used as the reporter antibody (Sigma-Aldrich, Bromborough, UK). An internal control and standard consisting of human full-length TFPI was provided by Baxter Innovations (Vienna, Austria). Absorbance was read at 450 nm on a plate reader (BioTek).
Von Willebrand factor ELISA
---------------------------
Von Willebrand factor (VWF) antigen (VWF:Ag) was measured by an ELISA technique. A 96-well DynexImmulon 4HBX plate (Fisher Scientific, Loughborough, UK) was coated with a capture antibody consisting of an antihuman VWF antibody from Dako (Ely, UK) and incubated at 4^o^C overnight. The plate was then blocked using a solution of 2% polyvinyl pyrrolidine. After washing the plate three times (wash buffer consisting of phosphate buffered saline and 0.05% Tween 20 v/v), 100 μl of test plasma and control plasma (CryoCheck Abnormal 1 and Abnormal 2 control plasma, Precision Biologic, Dartmouth, Nova Scotia, Canada) were added to the appropriate number of wells. A standard curve was constructed using Technoclone reference plasma from Pathway Diagnostics, UK. The plate was then covered and incubated for 2 h at room temperature. The plate was then washed three times and a rabbit anti-VWF horse-radish peroxidase immunoconjugate antibody (Dako) was added to each well. The plate was then covered and incubated for 1 h at room temperature before washing a further three times. Hundred microlitres of 3,3′,5,5′-tetramethylbenzidine liquid reagent (Skybio Ltd, Wyboston, Bedfordshire, UK) was added to each well and after 7 min 50 μl of Red-Stop reagent (Skybio Ltd) was added. Light absorbance was read at 405 nm on a plate reader (BioTek).
Thrombin generation assays
--------------------------
Whole blood samples were taken into vials containing 3.2% trisodium citrate with 20 μg/ml (final concentration in whole blood) corn trypsin inhibitor (CTI) at a ratio of one part anticoagulant to nine parts whole blood v/v (CTI; Cambridge Bioscience, Cambridge, UK), before heparin was given and 30 min after reversal of heparin by protamine sulphate. Platelet-poor plasma (PPP) was prepared by centrifuging samples twice at 1650 *g* and freezing in aliquots at −80°C for testing later. Thrombin generation was measured using the method described by Hemker *et al.*\[[@R6]\]. Experiments were performed by mixing 20 μl of commercially available trigger solution (Stago PPP Low; Diagnostica-Stago) with 80 μl of plasma in a U-bottom Immulon 2Hb 96-well plate (Diagnostica-Stago). The plate was maintained at 37°C using a plate incubator before being placed in a Fluoroscan Ascent fluorometer (Thermo Scientific, Helsinki, Finland). Thrombin generation was initiated by adding Flu-Ca (fluorogenic substrate Z-Gly-Gly-Arg; Bachem, UK, and 0.1 mol/l calcium chloride in a HEPES buffer). Fluorescence was measured using a Fluoroscan reader. Data were evaluated using Thrombinoscope software version 5.0.0.742 (Synapse BV, Maastricht, the Netherlands).
Calculation of concentrations of FFP, rFVIIa, PCC and anti-TFPI antibody to add
-------------------------------------------------------------------------------
Thrombin generation was measured in the presence and absence of a polyclonal anti-TFPI antibody (AF2974; R&D Systems, Abingdon, UK) at a concentration of 100 nmol/l, this concentration having previously been determined to fully neutralize full-length TFPI from physiologic concentrations (∼ 0.4 nmol/l) up to 6 nmol/l (data not shown).
Fresh frozen plasma
-------------------
Fresh frozen plasma was obtained from the Welsh Blood Service. The postoperative plasma volume for each patient was estimated using the formula, 0.07 × weight × (1 -- haematocrit). The mean plasma volume was then calculated for the whole cohort. Mean weight was also calculated. These figures were then used to calculate the volume of FFP to add to each sample to be equivalent to 15, 20 and 50 ml/kg. This was 0.308 ml FFP, 0.411 ml FFP and 0.617 ml FFP per ml patient plasma, respectively.
rFVIIa and PCC
--------------
The mean weight of the overall cohort was used to calculate the amount of rFVIIa to add to plasma to be equivalent to doses of 45, 90 and 180 μg/kg. rFVIIa was purchased from Novonordisk (NovoSeven, Novonordisk, Denmark). The final concentrations of rFVIIa in the spiked plasma were 0.93, 1.85 and 3.7 μg/ml, respectively.
The mean weight of the overall cohort was also used to calculate the amount of PCC to add to plasma equivalent to doses of 25, 35 and 50 U/kg. PCC was a gift from CSL Behring (Beriplex, CSL; Behring UK Ltd, Haywards Heath, UK). The final concentrations of PCC in the spiked plasma were 0.51, 0.72 and 1.03 U/ml, respectively.
Statistical analysis
--------------------
Data were analyzed using PASW Statistics version 18 software (SPSS Inc, released 2009; PASW Statistics for Windows, Version 18.0; Chicago, Illinois, USA). The Mann--Whitney *U* test was applied to examine differences between unrelated variables, whereas the Wilcoxon rank test was used to examine differences between related samples. Analysis of multiple related samples was performed using Friedman\'s test. Spearman\'s correlation coefficients were calculated to investigate the relationship between full-length TFPI and thrombin generation parameters.
Results
=======
A total of 102 patients were recruited, median age 68 (range 28--88), male 72, female 30. Ten patients were excluded from further analysis because they had heparin anti-Xa levels greater than 0.3 anti-Xa U/ml in the post-CPB samples. This left 92 patients in the analysis. Patients who were on warfarin were included provided their international normalized ratio before surgery was less than 1.5 as per institutional protocols.
Demographic and clinical parameters are summarized in Table [1](#T1){ref-type="table"}. Thrombin generation following CPB was lower compared with preoperative samples. Procoagulant factors and anticoagulants decreased significantly, apart from total TFPI, which increased significantly (median 55 ng/ml pre to 111 ng/ml post, *P* \< 0.0001) and VWF:Ag, which increased from 144 to 170 IU/dl (*P* \< 0.0001). The data are summarized in Table [2](#T2){ref-type="table"}.
CAT results are summarized for peak thrombin in Fig. [1](#F1){ref-type="fig"}, endogenous thrombin potential (ETP) in Fig. [2](#F2){ref-type="fig"}, lag time in Fig. [3](#F3){ref-type="fig"} and velocity index in Fig. [4](#F4){ref-type="fig"}. There was a fall in peak thrombin, ETP and velocity index in the post-CPB samples compared with pre-CPB samples, whereas lag time increased.
{#F1}
{#F2}
{#F3}
{#F4}
The addition of FFP resulted in a progressive, concentration-dependent increase in peak thrombin, ETP and velocity index and shortening of the lag time. The largest effect was seen with a concentration equivalent to 30 ml/kg (Figs. [1](#F1){ref-type="fig"}--[4](#F4){ref-type="fig"}). FFP at 15 ml/kg resulted in an increased peak thrombin to a level significantly greater than both preoperative and postoperative levels. Exceeding this volume resulted in a further dose-dependent increase. With rFVIIa, 45 μg/kg was sufficient to increase peak thrombin to preoperative levels, but not above preoperative levels, and exceeding this dose produced no further improvement. PCC at 25 U/kg resulted in peak thrombin greater than that seen preoperatively. There was no increase in peak thrombin between 25 and 35 U/kg, but there was a small but statistically significant difference comparing doses of 25 and 50 U/kg (*P* \< 0.0001). Inhibition of TFPI resulted in a peak thrombin concentration that was much higher than the preoperative level. Results for ETP were similar to peak thrombin.
FFP at a dose of 15 ml/kg was sufficient to correct the velocity index to preoperative levels. Exceeding this dose resulted in increases statistically significantly greater than preoperative levels. For rFVIIa and PCC, 45 μg/kg and 25 U/kg, respectively, were sufficient to return the velocity index to preoperative levels. Exceeding these doses produced no further increase. Inhibition of TFPI dramatically increased the velocity index well exceeding the preoperative level. Similar results were obtained for the lag time with the exception that it required 30 ml/kg of FFP before the lag time corrected to preoperative levels.
Full-length TFPI measured post-CPB was inversely correlated with ETP, peak thrombin and velocity index in the absence and presence of all concentrations of FFP, rFVIIa and PCC, the weakest correlation being seen with FFP. Thrombin generation in the presence of the anti-TFPI antibody did not correlate the full-length TFPI concentration. Figure [5](#F5){ref-type="fig"} summarizes these findings for ETP. Results for peak thrombin and velocity index were similar (data not shown).
{#F5}
Discussion
==========
The results show that 15 ml/kg of FFP was sufficient to correct ETP, peak thrombin and velocity index to preoperative levels. Larger volumes of FFP improved thrombin generation above baseline. rFVIIa at a dose of 45 μg/kg and PCC at a dose of 25 U/kg were sufficient to correct all measured thrombin-generation parameters to preoperative levels. Inhibition of TFPI markedly increased the ETP, peak thrombin and velocity index beyond preoperative levels and shortened the lag time.
FFP remains the main treatment available to most clinicians for correcting any coagulopathy. The optimum dose of FFP to return the majority of thrombin-generation parameters to preoperative levels was 15 ml/kg. Some studies have reported a correlation between lag time and bleeding \[[@R26]\] and 30 ml/kg was required to correct this. The full-length TFPI concentration correlated with all thrombin-generation parameters postoperatively and remained so in the presence of FFP, rFVIIa and PCC. This supports the view that TFPI is an important determinant of thrombin generation and similar findings have been reported by Knappe *et al.*\[[@R22]\] and Peraramelli *et al.*\[[@R27]\]. The inverse correlation of full-length TFPI concentration with thrombin-generation parameters was weaker following the addition of FFP than after adding rFVIIa and PCC. This may be because FFP contains both procoagulant factors and inhibitors and would tend to normalize both, whereas rFVIIa and PCC only affect the procoagulant pathways. There was no correlation between full-length TFPI and thrombin generation in the presence of an anti-TFPI antibody, further supporting the importance of TFPI in thrombin generation assays activated with low concentrations of tissue factor.
The enhancing effect of FFP observed on thrombin generation may be overestimated by the in-vitro nature of the assay. Patient plasma was contact-inactivated by addition of CTI and so after addition of rFVIIa, PCC and TFPI antibody thrombin generation was activated through tissue factor. In contrast, the added FFP was not CTI-inhibited and would therefore have been prone to contact activation in addition to tissue factor activation. Contact activation has previously been shown to result in higher ETP and peak thrombin and shorter lag time when using lower concentrations of tissue factor as used in this study \[[@R28],[@R29]\].
rFVIIa made no significant difference to thrombin-generation parameters beyond doses of 45 μg/kg, indicating that lower doses are as effective as higher doses in enhancing thrombin generation *in vitro*. This is as would be predicted given the global reduction in clotting factors observed in this patient cohort. A similar result was observed by Altman *et al.*\[[@R30]\] using a model of dilutional coagulopathy. This suggests that rFVIIa may be inappropriate to use in isolation to treat bleeding. Although there are a number of studies that have reported safety using rFVIIa in paediatric patients undergoing CPB \[[@R31]\], some studies have reported an increase in thrombotic events in adult patients \[[@R32]\]. Studies looking at adult patients from a broader population have also shown an increase in arterial events \[[@R33]\], something that would be particularly deleterious in the typical adult patient population undergoing cardiac surgery. Therefore, based on these in-vitro assays, if rFVIIa is used, lower doses may offer the best balance of achieving a beneficial effect while possibly reducing the chance of an adverse event.
PCC corrected thrombin-generation parameters to preoperative levels at a concentration of 25 U/kg. There was some further statistically significant improvement at concentrations of 50 U/kg; although this was so small, it is unlikely to be clinically significant. This finding is consistent with other in-vitro studies in which PCC has been shown to enhance thrombin generation \[[@R34]\]. In-vivo PCC has been shown to be effective in reducing blood loss in human studies \[[@R16],[@R35]\]. As well as different numbers of clotting factors, different PCCs contain different amounts of heparin, with some containing more heparin than the PCC used in this study, which in turn has been shown to affect thrombin generation \[[@R36]\]. Although a four-factor PCC was used in this study, others have shown similar effects on thrombin generation using a three-factor PCC \[[@R34]\]. PCC has the advantage of additional safety over FFP for transmission of infection and smaller volumes in patients with cardiac compromise. However, large doses have been associated with disseminated intravascular coagulation in some studies \[[@R18]\].
Inhibition of TFPI had a marked effect on enhancing thrombin generation. TFPI *in vivo* is increased by administration of heparin \[[@R19]\]; the anticoagulant is routinely used in CPB and therefore may be a good therapeutic target in this group of patients. TFPI inhibition has previously been reported to enhance thrombin generation in plasma from people with haemophilia \[[@R21],[@R37]\]. However, a number of studies have reported higher ETP and peak thrombin concentration in patients with prothrombotic tendencies \[[@R38]--[@R40]\]. This suggests caution may be required if TFPI inhibition was to be used as a therapeutic target especially in cardiac patients in which arterial thrombosis is often already a significant risk.
A weakness of studies investigating thrombin generation in plasma is that they take no account of other important factors that require correction in controlling bleeding following cardiac surgery: fibrinogen concentration, platelet count and function and fibrinolysis. A low fibrinogen has been associated with excessive postoperative bleeding in a number of studies \[[@R41],[@R42]\]. Thrombocytopenia has also been previously described as a risk factor for bleeding following CPB \[[@R8],[@R43]\]. Clot formation and durability require sufficient thrombin to cleave fibrinogen to fibrin, FXIII to cross-link fibrin monomers, platelets and reduced fibrinolysis. Therefore, correction of thrombin generation forms only one component in the management of bleeding.
In summary, this study suggests that comparatively low doses of FFP, rFVIIa and PCC may be sufficient to correct thrombin generation in patients who have undergone surgery requiring CPB. Inhibition of TFPI may offer a future therapeutic strategy for managing bleeding in this group of patients. However, these in-vitro results need to be validated *in vivo*.
Acknowledgements
================
This work was funded by the British Heart Foundation award reference FS/11/42/28753. R.H., M.D. and F.S. are employees of Baxter Innovations GmbH.
C.L.P. designed the study, recruited the patients, performed the laboratory assays, interpreted data and wrote the article. R.J. assisted in recruitment of patients, sample preparation and edited the article. R.H., M.D. and F.S. designed some of the laboratory studies, interpreted data and edited the article. S.B. and D.M. assisted in the design of the study, recruitment of patients and edited the article. J.E.H. and V.B.O. designed the study, planned experiments, interpreted data and edited the article. P.W.C. designed and supervised all aspects of the study and edited the article.
Conflicts of interest
---------------------
There are no conflicts of interest.
######
Demographic details and clinical parameters of study participants
Parameter Data
---------------------------------- --------------------
Age (years) 67
28--88 (62--76)
Weight (kg) 76
48--134 (67--90)
Sex (M/F) 64/28
Time on bypass (min) 146
64--427 (117--193)
Aortic cross-clamp time (min) 115
50--34 (96--152)
Intraoperative 14
volume crystalloid (ml/kg) 0--36 (8--20)
Intraoperative volume of colloid 13
(ml/kg) 0--47 (8--18)
Cell salvage volume (ml/kg) 10
3--52 (7--12)
Data shown are median, range (interquartile range).
######
Concentrations of procoagulant factors, inhibitors, platelet count and thrombin-generation parameters before and after CPB
Parameter Pre-CPB Post-CPB Comparison between pre-CPB and post-CPB (*P* value)
--------------------------------- --------------------- ------------------------- -----------------------------------------------------
Platelet count (×10^9^ cells/l) 223 115 \<0.0001
124--670 (118--257) 42--258 (98--138)
APTT (s) 29 31 \<0.0001
23--38 (28--31) 22--46 (30--35)
PT (s) 11 15 \<0.0001
10--16 (10--12) 12--22 (14--17)
vWF antigen (IU/dl) 144 170 \<0.0001
61--322 (113--184) 84-367 (131--205)
Fibrinogen (g/l) 3.2 1.6 \<0.0001
1.7--7.9 (2.6--4.0) 1.1--3.8 (1.4--1.9)
Factor II (IU/dl) 95 52 \<0.0001
44--152 (85--107) 33--83 (47--60)
Factor V (IU/dl) 92 50 \<0.0001
51--135 (76--109) 17--93 (40--58)
Factor VII (IU/dl) 104 68 \<0.0001
48--192 (80--118) 25--146 (58--80)
Factor VIII (IU/dl) 129 92 \<0.0001
75--262 (106--157) 49--216 (74--117)
Factor IX (IU/dl) 133 104 \<0.0001
59--224 (117--151) 62--157 (93--123)
Factor X (IU/dl) 90 46 \<0.0001
27--148 (78--104) 20--57 (38--54)
Factor XI (IU/dl) 95 60 \<0.0001
42--183 (82--109) 35--137 (50--72)
Factor XIII (%) 103 57 \<0.0001
30--213 (89--117) 21--107 (45--68)
Full-length TFPI (ng/ml) 21 15 \<0.0001
7--39 (16--24) 4--35 (11--22)
Total TFPI (ng/ml) 56 113 \<0.0001
32--99 (48--67) 60--165 (93--127)
Anti-thrombin (%) 92 51 \<0.0001
11--144 (83--100) 30--77 (45--58)
Free protein S (%) 84 46 \<0.0001
43--125 (72--98) 30--85 (39--55)
Protein C (%) 106 61 \<0.0001
64--204 (92--119) 39--93 (52--69)
Lag time (min) 7 9 0.026
3--18 (6--8) 0--36 (7--14)
Peak thrombin (nmol/l) 35 11 0.019
1--115 (21--52) 0--144 (2--42)
ETP (nmol/l per min) 339 155 \<0.0001
10--867 (221--497) 0--1161 (0--376)
Velocity index (nmol/l per min) 9 2 \<0.0001
2--84 (5--14) 0--53 (0--9)
Anti-Xa activity (IU/ml) -- 0.12 --
0.00--0.30 (0.08--0.19)
Data shown are median and range (interquartile range). APTT, activated partial thromboplastin time; CPB, cardiopulmonary bypass; ETP, endogenous thrombin potential; PT, prothrombin time; TFPI, tissue factor pathway inhibitor; vWF, von Willebrand factor.
| {
"pile_set_name": "PubMed Central"
} |
Heat stress is an environmental and occupational hazard. The prevention of deaths caused by extreme high temperatures (above 42 °C) is an issue of human public health importance and also a concern in animal production[@b1]. In 1995, a heat wave in Chicago resulted in 700 reported deaths, of which 39% occurred in individuals known to have a prior heart condition[@b2]. Similar reports indicated extreme temperatures caused a large number of deaths and heat-related injuries in Europe in 2003 and in California in 2006[@b3][@b4]. Mammals have sweat glands, but can still be affected by high temperatures and even suffer sudden death. The conception rates of lactating Brown Swiss, Jersey and Holstein cows in the United States were reported to decrease from 52% to 32% as the maximum air temperature increased from 23.9 °C to 32.2 °C during the summer[@b5]. In Brazil, the pregnancy rates of Holstein cows in free stall reduced from 71.2% in the winter to 45.7% in the summer[@b6].
High temperatures can also harm poultry, especially broiler chickens, which have no sweat glands and poor thermotolerance[@b7]. Heat stress causes a series of physiological and metabolic changes in broiler chickens, including elevated body temperature, panting, respiratory alkalosis and sudden death[@b8]. According to our previous *in vivo* and *in vitro* research in rats, sudden death caused by heat stress (41 ± 1 °C) is primarily the result of pathological changes in the heart, such as necrosis and cell degeneration[@b9]. The heart is a critically important organ in animals and humans. Previous clinical reports demonstrated that thermal tolerance to heat stress is impaired in human patients with cardiovascular disease. In particular, patients with cardiovascular conditions associated with ventricular dysfunction and chronic heart failure (CHF) are predisposed to heat intolerance[@b10][@b11]. Despite the fact that high heat stress-induced mortality rates of nearly 40% have been reported among broiler chickens exposed to a temperature of 40 °C, little attention has been paid to investigation of the mechanisms associated with heat stress in broiler chickens.
The heat shock response (HSR) is an evolutionarily-conserved defense mechanism that protects against sudden stresses, such as elevated temperature or environmental changes[@b12]. Heat shock proteins (HSP) are a family of proteins that are produced by cells in response to exposure to stressful conditions. They were first described in relation to heat shock[@b13], but are now known to also be expressed during other stresses including exposure to cold[@b14], UV light[@b15], and during wound healing or tissue remodeling[@b16]. Many members of this group perform chaperone function by stabilizing new proteins to ensure correct folding or by helping to refold proteins that were damaged by the cell stress[@b17]. HSPs are found in virtually all living organisms, from bacteria to humans. Heat-shock proteins are named according to their molecular weight. For example, Hsp60, Hsp70 and Hsp90 (the most widely studied HSPs) refer to families of heat shock proteins on the order of 60, 70, and 90 kilodaltons in size, respectively[@b18]. Some previously researches transfected Hsp90, Hsp60 and Hsp56 into rat H~9~C~2~ cells and cultured primary cardiac cells to investigate the mechanisms related to stress responses. Interestingly, these experiments clearly indicated that overexpression of Hsp90 protected against a number of stimuli[@b19][@b20]. Our other studies[@b21] indicated overexpression of Hsp70 exerted a protective effect against heat stress[@b22][@b23]. Therefore, these studies indicate the individual Hsps exert specific protective effects and their functions should be assessed separately. However, the protective roles of Hsp60 and other Hsps have not been fully investigated[@b24][@b25].
The small heat shock protein αB-Crystallin (CryAB, HSPB5, 20 KDa) belongs to the small heat shock protein (sHsp) family, and is expressed in most organisms in response to several types of stress (e.g., UV, hyperthermia, toxic radicals) in order to protect cells[@b26][@b27]. CryAB is expressed ubiquitously throughout the mammals' body and exerts a variety of highly protective functions to maintain homeostasis. CryAB has also been shown to exert anti-apoptotic properties, as it can prevent cell death in response to conditions such as stroke[@b28] by maintaining the cell cytoskeleton[@b29]. Mechanistically, CryAB localizes to the I-band and M-line region in myofibrils and has been confirmed to play a myofibril-stabilizing role in cardiomyocytes *in vitro*[@b30]. It is now believed that the pleotropic functions of CryAB are the result of its diverse interactions with a wide variety of different proteins[@b31]. In our previous research, we investigated CryAB expression and localization, and demonstrated it plays a protective role against heat stress in rat myocardial cells *in vivo* and *in vitro*[@b21]. However, the function of CryAB remains poorly explored compared to other members of the sHsp family.
Aspirin (acetylsalicylic acid, ASA) is widely used as a drug to treat pain, fever and inflammation[@b32] and is also prescribed long-term at low doses to help prevent heart attacks, strokes and blood clot formation in high risk individuals[@b33]. Therefore, ASA is an important part of the care of patients who have suffered a myocardial infarction (heart attack)[@b34]. ASA has been reported to reduce the development, growth rate, or both, of several human cancers in animal models[@b35], via a mechanism mediated at least in part by inhibition of the cyclo-oxygenase (COX) enzymes and reduced production of prostaglandins and other inflammatory mediators[@b36]. Interestingly, heat stress has been shown to active multiple cellular responses, including inflammatory pathways[@b35].
In present study, we established *in vivo* and *in vitro* models to investigate the functions on the chicken myocardium during prolonged heat stress. Moreover, we also assessed the ability of ASA to protect the myocardium against heats stress *in vivo* and *in vitr*o, and investigated the relationship between CryAB and ASA during heat stress.
Materials and Methods
=====================
Establishment of the *in vivo* heat stress model
------------------------------------------------
All experiments were performed in accordance with the guidelines of the Animal Ethics Committee of Jiangsu Province (China) and were approved by the Institutional Animal Care and Use Committee of Nanjing Agricultural University, China. The *in vivo* heat stress model was established as previously described[@b37]. Briefly, one-day-old specific pathogen free (SPF) chickens were purchased from Qian Yuan Hao Biotechnology Company, Nanjing, China. The entire population was vaccinated against Newcastle disease (ND) and infectious bursal disease (IBD) on day 7 and day 14, respectively. The birds were allowed to acclimate to their new housing and recover from environmental stress for 30 days. Then, 270 chickens were randomly divided into three groups, designated the HS group (heat stress), the ASA(−)HS(+) group (pretreated with aspirin before heat stress), and the ASA(+)HS(−) group (pretreated with aspirin, no heat stress), respectively, with 90 chickens in each group. The chickens were not provided water for the 12 h leading up to the experiment. The chickens in the ASA(+)HS(−) and ASA(+)HS(+) groups were administered aspirin orally at 1 mg/kg body weight (ASA powder \>98% purity; Sigma, USA) 2 h before the heat stress phase of the experiment. During the heat stress phase, the chickens in ASA(+)HS(−) group were housed under normal conditions as a control group (i.e. not exposed to heat), while the chickens in the ASA(+)HS(+) and ASA(−)HS(+) groups were exposed to heat stress by rapidly, gently moving the animals from 25 ± 1 °C to an air chamber (GJ-1, Suzhou Fengshi Laboratory Animal Equipment Co. Ltd, China) preheated to 40 ± 1 °C at 60\~70% humidity. The broilers were allowed free access to food and water *ad libitum* during heat stress exposure. At 0, 1, 2, 3, 5, 7, 10, 15 and 24 h, 10 chickens in each group were sacrificed humanely by decapitation. The body temperature of the broilers was measured via the rectum (3 cm inside) using a mercury thermometer within 2 min before slaughter. The hearts were excised and fixed in 10% formalin for pathological studies or frozen in liquid nitrogen for ELISA.
For pathological analysis, heart samples from the 10 chickens in each treatment group were fixed in 4% methanol, paraffin-embedded, cut into 4 μm-thick serial sections, stained with hematoxylin (3 min) and eosin (1 min), and examined by light microscopy (Imager A2, ZEISS, Germany).
Isolation and culture of primary chicken myocardial cells
---------------------------------------------------------
All experiments were performed in accordance with the guidelines of the Animal Ethics Committee of Jiangsu Province (China) and were approved by the Institutional Animal Care and Use Committee of Nanjing Agricultural University, China. Twelve-day-old specific pathogen free (SPF) embryonated eggs (Qian Yuan Hao Biotechnology Company, Nan Jing, China) were opened. The hearts of the embryos were removed in a bio-clean environment, cut into pieces (1 mm^3^), washed in pre-cooled PBS (4 °C) four times, the fibrous tissue was digested in 1 mg/mL collagenase type I (17100-017, Gibco, USA) at 4 °C for 14--16 h, then the digestion was terminated by addition of Dulbecco's modified eagle medium (DMEM; 11995-065, Gibco) containing 20% fetal bovine serum (FBS; 16141079, Gibco). The digest was centrifuged at 1000 rpm for 10 min, the cells were resuspended in DMEM containing 20% FBS, 100 units/mL penicillin and 100 units/mL streptomycin. The cells were cultured in cell culture plates at 37 °C in a humidified atmosphere of 5% CO~2~ and 95% air, then the primmorph were transferred into new cell culture plates after 1 h and 0.1 mmol/L 5-Bromo-2′-deoxyuridine solution (BRDU, B16880, Sigma, Germany) was added the culture solution. Cells were cultivated for 48 h to let them adhere.
Immunofluorescence analysis was employed to confirm the cells were myocardial cells using the marker alpha actinin (α-actinin, ab11007, abcam, USA) which is specific to myocardial cells. The proportion of myocardial cells was assessed in five randomly chosen fields of view (data not shown) by fluorescence microscopy (Cx41-32rfl, Olympus, Japan).
Assessment of cell viability *in vitro*
---------------------------------------
Four groups of primary chicken cells were established (HS: 2 h, ASA + HS: 2 h, HS: 24 h, ASA + HS: 24 h). Cells (1 × 10^6^) were seeded and cultured in 6 cm cell culture plates (Corning, USA), and used in experiments when the cell division was higher than 85%. After treatment with ASA and/or HS, the cells were digested with 0.25% trypsin for 10 min, resupsended in DMEM containing 20% FBS and 1% penicillin/streptomycin, then counted using the Trypan blue assay under an inverted light microscope. Each sample was counted three times.
CryAB ELISA
-----------
The chicken heart samples were washed in ice-cold saline and homogenized on ice in 10 volumes of homogenization buffer \[0.15 M NaCl, 20 mM Tris-HCl (pH 8.0), 1 mM ethylenediaminetetraacetic acid, 1 mM phenylmethylsulphonyl fluoride, 0.1 μM E-46, 0.08 μM aprotinin, 0.1 μM leupeptin, and 0.1% NP-40\] using an Ultra-Turrax homogenizer (623003, Fluko Equipment Shanghai Co. Ltd, China) The homogenates were centrifuged at 12,000 g for 20 min at 4 °C to remove debris, and the supernatant was collected and stored at −20 °C for protein quantification.
Primary chicken myocardial cells were washed twice with PBS and lysed in M-PERH mammalian protein extraction reagent (28501, Thermo Scientific, Waltham, MA, USA) supplemented with Halt^TM^ protease inhibitor cocktail (78425, Thermo Scientific, Waltham, MA, USA) according to the manufacturer's instructions, centrifuged at 14,000 g for 5 min at 4 °C and the supernatants were used as total protein extracts.
The protein concentrations of the samples were measured using a Micro-BCA™ protein assay kit (23235, Thermo Scientific, Waltham, MA, USA), then CryAB protein levels were quantified using a commercially available ELISA Kit (MBS2882479, MyBioSource, USA) according to the manufacturer's instructions.
Statistical analysis
--------------------
Curve Expert 1.3 software was used to generate the standard curves for the ELISAs. Data were compared with the baseline level (0 h in the HS group) by one-way analysis of variance (ANOVA) followed by Fisher's least significant difference (LSD) test using SPSS version 21 for Windows and Graphpad prism 6.0 software. *P *\< 0.05 was considered significant; *P *\< 0.01 was considered highly significant. Duncan's multiple range test was used to analyze CryAB levels between groups at each time point. All raw data presented are expressed as the mean ± standard deviation (SD). All experiments were repeated three times.
Results
=======
Clinical manifestation of heat stress in broiler chickens
---------------------------------------------------------
The broilers behaved normally after being orally administered ASA in water 2 h before heat stress. After 1 h heat stress, the broilers in the ASA(−)HS(+) group displayed polypnea and became more sensitivity to heat compared to the ASA(+)HS(+) pretreated group. After 5 h of heat stress, the broilers in the ASA(−)HS(+) group were not as active, and the ASA(+)HS(+) treatment group displayed similar to their behavior at earlier time point. From 10 h to 24 h of heat stress, the ASA(+)HS(+) group showed relaxed behavior compared to ASA(−)HS(+) group.
Both groups subjected to heat stress had higher rectal temperatures than the control group housed at RT (25 °C; [Fig. 1](#f1){ref-type="fig"}); however, the ASA(−)HS(+) group had significantly higher body temperatures compared to the ASA(+)HS(+) group at all time points during heat stress (P* *\< 0.05). The mercury thermometer used in this study ranged from 25 to 43 °C. During heat stress, the body temperature of most chickens in the ASA(−)HS(+) group was \>43 °C (we assumed a body temperature of 44 °C).
Histopathological changes in the myocardium in the *in vivo* model of heat stress
---------------------------------------------------------------------------------
The histopathological changes in the myocardium of the heat stressed groups and control group are shown in [Fig. 2](#f2){ref-type="fig"}. In the present experiment, some chickens were pre-treated with ASA 2 h before heat stress, euthanized at 0 h of heat stress and defined as the ASA control group ([Fig. 2F](#f2){ref-type="fig"}). As shown in [Fig. 2C](#f2){ref-type="fig"}, severe damage to the heart tissue was evident after 5 h of heat stress. The space between muscle fibers became wider (▲) and karyopyknosis could be observed (→). In the ASA(+)HS(+) group, cell swelling was observed, accompanied by slightly wider spacing between the heart fibers. After 15 h heat stress ([Fig. 2D](#f2){ref-type="fig"}), bleeding (←) and cell death were the major pathological changes. The heart tissue lost its normal organization and structure, with the cell cytoplasm fusing together, and nuclei disappearing or undergoing karyopyknosis. In the ASA(+)HS(+) group ([Fig. 2I](#f2){ref-type="fig"},G), the cardiac fibers became narrowed and chaotic after 15 h heat stress compared with the no heat stress control group, with fewer blood cells observed in the intercellular spaces. Most cells had degenerated, and a few were necrotic (→).
Effect of heat stress on the survival of primary myocardial cells *in vitro*
----------------------------------------------------------------------------
Primary chicken myocardial cells were treated with or without 1 mg/mL ASA ASA for 2 h, cultured at 42 °C for 2 h or 24 h, and cell numbers were determined ([Fig. 3](#f3){ref-type="fig"}). After 2 h of heat stress, there was significantly more cells in the ASA pretreated group than the HS group (*P *\< 0.01). After 24 h heat stress, the number of cells had reduced in both groups compared to the numbers after 2 h heat stress; however, significantly more cells had survived in the ASA pretreated group than the HS group after 24 h heat stress (*P *\< 0.01).
Effect of heat stress on CryAB protein expression *in vivo*
-----------------------------------------------------------
The expression of CryAB was quantified in the myocardium of the chickens using an ELISA ([Fig. 4](#f4){ref-type="fig"}). At 0 h heat stress, the concentration of CryAB in the myocardium of the control group was only 200 pg/mL compared to 1200 pg/mL in the groups pretreated with ASA for 2 h, which represents a 6-fold difference (*P *\< 0.01).
In the HS group, expression of CryAB increased after 1 h heat stress, reached the highest level after 10 h of heat stress (1400 pg/mL), then slightly decreased but still remained high between 15 and 24 h heat stress (1000 pg/mL). In the ASA group pretreated with ASA for 2 h, CryAB peaked at 0 h and 3 h (i.e., 2 and 5 h after administration of ASA), then sharply decreased after 3 h (p* *\< 0.01) and recovered to control levels (200 pg/mL) after 24 h heat stress. In the ASA(+)HS(+) group, the expression of CryAB followed a different trend. After 1 h of heat stress, CryAB was expressed at 6-fold lower levels compared to the HS group not pretreated with ASA, then slightly increased at 2 h, but remained lower than the levels in the ASA(−)HS (+) group up to the end of heat stress (24 h) with the exception of a peak at 7 hours (1000 pg/mL). At 3 h of heat stress, the level of CryAB was significantly lower in the ASA(+)HS(+) group than the HS group (p* *\< 0.01).
CryAB expression in primary chicken myocardial cells in the *in vitro* model of heat stress
-------------------------------------------------------------------------------------------
CryAB expression was also measured in primary chicken myocardial cells *in vitro* after different treatments ([Fig. 5)](#f5){ref-type="fig"}. Before heat stress, CryAB was expressed at significantly higher levels in ASA-pretreated cells than control cells that had not been treated with ASA (p* *\< 0.01). However, cells exposed to 1 h heat stress expressed significantly higher levels of CryAB compared to not only before HS (P* *\< 0.01), but also compared to the ASA(+)HS(−) and ASA(+)HS(+) groups after 1 h heat stress. The levels of CryAB remained higher in cells subjected to heat stress \[ASA(−)HS(+)\] until 15 h compared to control cells. After 24 h heat stress, CryAB expression returned to the same level as control cells before heat stress. In the ASA(+)HS(+) group, CryAB expression decreased after 1 h and remained at a low level until 24 h heat stress.
Discussion
==========
Heat stress is a non-specific stressor that can affect the welfare of livestock and even contribute to death. Our previous research confirmed short term exposure to heat stress (1 h) induced detectable levels of enzymes related to heart damage, such as AST, CK and CKMB, in both the serum of rat blood and supernatant of rat myocardial cells[@b9][@b38]. Furthermore, pathological lesions, mainly due to necrosis, were observed in the rat myocardium after 40 min of heat stress *in vivo*, accompanied by lower CryAB expression, indicating CryAB may play an important role to protect the mammalian heart against heat stress[@b21]. In the present research, we investigated the expression profile of CryAB in the chicken myocardium using *in vivo* and *in vitro* models of heat stress lasting up to 24 h. Furthermore, we assessed the ability of ASA to protect the myocardium during heat stress, as it has been identified that the drug may protect the heart from heat stress by inducing Hsp90 and HspB1 expression[@b39]. The aim of this study was to determine whether CryAB could be induced by ASA and protect the myocardium during heat stress in broiler chickens using *in vivo* and *in vitro* models.
In the *in vivo* model, heat stress resulted in severe damage (observed after 5 to 24 h) in the myocardium; these changes were mainly characterized by karyolysis and necrosis, signifying cell death. In comparison, the myocardium was less severely damaged in the ASA pretreated group, in which degeneration of cytoplasm was observed after 1--15 h of heat stress and the myocardium became disorganized with a number of necrotic and karyolitic cells after 24 h. The body temperature of the chickens in the ASA(+)HS(−) group remained approximately 41 °C, which is the normal body temperature range for broilers. Heat stress increased the body temperature of the chickens; however, the ASA(+)HS(+) group maintained a lower body temperature (41--43 °C) than the ASA(−)HS(+) group (42--44 °C). ASA is widely used to treat pain, fever and cardiac disease in humans[@b36]. During heat stress, body temperature increases and homeostasis of the intestinal flora becomes disrupted, which may cause inflammatory factors such as TNF-α, IL-6, IL-8, among others, to be released into the circulation, which can lead to fever. ASA can involve into anti-inflammatory properties[@b35]. In the *in vivo* model, ASA played a key role to protect the heart from heat stress. Aspirin controls fever by acting on the prostaglandin system as it irreversibly inhibits the cyclooxygenase (COX) enzymes[@b36]. Pathological analysis confirmed pretreatment with ASA attenuated heat stress-induced damage to the myocardium *in vivo* and prevented heat stress-induced primary chicken myocardial cell death *in vitro.* The ability of ASA to protect the myocardium during heat stress is consistent with the fact that low dose ASA is widely used to reduce the risk of subsequent heart attacks and prevent the death of myocardial tissue[@b40].
CryAB is a member of the small heat shock protein family (sHSPs, HspB1-HspB10) that is expressed in several organs and functions as a major cytoskeletal chaperone protein. The cytoskeleton plays a key role to maintain the physical shape of heart cells, protect against endoplasmic reticulum (ER) stress, and preserve mitochondrial function *in vivo* and *in vitro*. Mitochondria consume oxygen and provide ATP throughout the entire body, and occupy up to 60% of the volume of myocardial cells[@b41][@b42]. Mammalian CryAB expression and localization have been well-characterized both *in vivo* and *in vitro* in the rat[@b9][@b21]; these analyses indicated CryAB may play an important role in response to heat stress in cardiac cells. However, until now, the expression of CryAB had not been researched in poultry. Our *in vivo* expression profiling of broiler chickens revealed 2 h pretreatment with ASA increased the expression of CryAB in the myocardium by 6-fold compared to control chickens. CryAB reached the highest level after 2 h heat stress; however, CryAB expression decreased significantly (by nearly 4-fold) after 3 h heat stress in chickens pretreated with ASA. After 24 h heat stress, CryAB expression recovered to similar levels as untreated control animals (200 pg/mL) in the chickens pretreated with ASA. Exposure to heat stress in the absence of ASA resulted in the opposite trend: CryAB expression increased slightly after 1 h heat stress, reached the highest level after 10 h heat stress and remained high up to 24 h heat stress. These results suggest that pretreatment with ASA induced the expression of CryAB during the early stage of heat stress. This reflects the metabolism of ASA; the half-life of ASA in humans is approximately 2.0 to 4.5 h. In our *in vivo* model, the chickens were orally administered ASA 2 h before heat stress.
Pathological analysis of the *in vivo* model further confirmed that pretreatment with ASA to induce CryAB expression played a critical role to protect the myocardium against heat stress. In the ASA(+)HS(+) group, the expression of CryAB was strongly upregulated by pretreatment with ASA, slowly decreased during the initial stage of heat stress (1 h), slowly increased between 2 and 7 h heat stress, and then returned to similar levels as the HS group. CryAB has been shown to bind to other cytoskeletal proteins such as desmin and vimentin, which in turn can decrease CryAB expression[@b28][@b43]. The expression profiling suggested ASA only induced CryAB expression at the early stage (1--4 h) of heat stress.
The expression of CryAB was not the same *in vivo* and *in vitro* in primary chicken myocardial cells. CryAB did not dramatically increase at the beginning of heat stress (0--1 h) and did not sharply decrease after prolonged heat stress. Moreover, the heat stressed cells expressed higher levels of CryAB at several time points (1, 3, 5, 15 h) than ASA(+)HS(+) cells. In contrast to the *in vivo* model, ASA treatment induced lower levels of CryAB than HS *in vitro*. Unlike other Hsps (Hsp90, Hsp27) we have researched[@b39][@b44], it may not possible to induce overexpression of CryAB *in vitro*. As previously stated, aspirin controls fever via the prostaglandin system by irreversibly inhibiting the cycloxygenases. However, cells in culture do not possess a hormone regulatory system. The main functional region of CryAB is its αB-crystallin chain, which means it cannot be induced by acid or other stressors like the inducible Hsps. However, CryAB could not be induced by pretreatment with ASA in primary chicken myocardial cells in the present study, which could reflect specific differences between the mammalian and chicken prostaglandin systems. ASA can induce Hsf1 expression, which binds to heat shock element (HSE) to induce the transcription of Hsp genes. However, unlike Hsp90, CryAB was not identified as a major Hsp that can be induced by Hsf1[@b45][@b46]. Similarly, CryAB showed no significant expression changes in primary chicken myocardial cells treated with ASA. Therefore, the mechanism by which ASA induces the expression of CryAB *in vivo* still needs to be identified. In conclusion, this study indicates that ASA can induce expression of CryAB and protect the myocardium *in vivo,* but not *in vitro*.
Additional Information
======================
**How to cite this article**: Tang, S. *et al*. Aspirin upregulates αB-Crystallin to protect the myocardium against heat stress in broiler chickens. *Sci. Rep.* **6**, 37273; doi: 10.1038/srep37273 (2016).
**Publisher's note**: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The current study was supported by grants from the National Key Basic Research Program of China (973 Program; grant no. 2014CB138502), the National Natural Science Foundation of China (grant no. 31602027), the National Natural Science Foundation of China (grant no. 31672520), the National Natural Science Foundation Of China (grant no. 31372403), Jiangsu Natural Science Foundation of China (grant no. BK20160732),China Postdoctoral Science Foundation(2016M591860), the Priority Academic Program Development of Jiangsu Higher Education Institutions, Graduate Research and Innovation Projects in Jiangsu Province and the Sino-German Agricultural Cooperation Project of the Federal Ministry of Food, Agriculture and Consumer Production, Berlin, Germany.
**Author Contributions** S.T. Designed the experiment and wrote this manuscript and Figure 4. B.Y. helped to revise the manuscript. E.B.S. organized original data. H.B.C. contributed to the *in vivo* experiment and prepared Figures 1 and 2. Y.F.C. prepared Figure 3. X.H.Z. contributed to Figure 5. E.D.B. is corresponding author. J.H. revised the language of the manuscript. All authors reviewed the manuscript.
{#f1}
{#f2}
{#f3}
{#f4}
{#f5}
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .601}
{#sp2 .602}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
*Culicoides* biting midges (Diptera: Ceratopognidae) are vectors of viruses of both medical and veterinary importance (Purse and Venter [@CR39]). African horse sickness virus (AHSV), epizootic haemorrhagic disease virus (EHDV) and bluetongue virus (BTV) are listed by the Office International des Epizooties as posing a high risk to animal health where competent vectors are present. The emergence of Schmallenberg disease in Europe has further highlighted the importance of these insects in transmitting new zoonoses endangering animal welfare (Carpenter et al. [@CR16]). Even in the absence of disease, the pain and distress of *Culicoides* blood feeding causes major economic losses (Mordue and Mordue [@CR34]), and can result in life-threatening allergic reactions (Carpenter et al. [@CR14]).
*Culicoides impunctatus* is geographically one of the most widespread pest species of biting midge across the West Palaearctic region (Mathieu et al. [@CR33]). The species has a broad host range, including both wildlife and livestock (Blackwell et al. [@CR8]), and causes economic losses to tourism and forestry through voracious blood feeding on people (Hendry and Godwin [@CR25]). Female *C. impunctatus* are autogenous, producing one batch of eggs prior to a blood meal, facilitating population growth up to huge densities even where hosts are not readily available (Blackwell et al. [@CR6]; Boorman and Goddard [@CR10]). In addition to being a serious economic pest, *C. impunctatus* is also susceptible to infection by BTV, as determined under laboratory conditions (Carpenter et al. [@CR13]). Given the high population densities and biting rates that can be attained by *C. impunctatus*, the potential of this species to act as a vector of disease cannot be entirely discounted (Purse et al. [@CR40]).
Currently there are few effective tools for the surveillance and control of biting midge populations (Carpenter et al. [@CR14]). Traps baited with carbon dioxide (CO~2~) and 1-octen-3-ol, kairomones emitted by mammalian hosts, have been tested and marketed for controlling populations of biting midge, including *C. impunctatus* (Mands et al. [@CR32]). While successful in capturing biting midges, these traps were originally designed for catching mosquitoes, and may not be fully optimized for use against *C. impunctatus* and other biting midge species. Laboratory and field studies have demonstrated that *C. impunctatus* responds to a wide range of host-produced kairomones, including acetone, butanone, lactic acid and a number of phenolic compounds (Bhasin et al. [@CR4]; Logan et al. [@CR31]). The addition of cow urine and hexane extracts of hair samples from large animals have also been shown to increase the attraction of *C. impunctatus* to traps baited with CO~2~ alone, and CO~2~ and 1-octen-3-ol, respectively (Bhasin et al. [@CR5]; Mands et al. [@CR32]). While the chemical components responsible for this increase in attraction of *C. impunctatus* to animal odor have not been elucidated fully, in analyses by coupled gas chromatography and electroantennographic detection (GC-EAD) in our laboratory volatile components have been identified from cow urine and hair that elicit antennal responses in a related species, *C. nubeculosus* (Isberg et al. [@CR27]). In a laboratory behavioral assay, 1-octen-3-ol, decanal, 2-ethylhexanol, phenol and 4-methylphenol elicited attraction of *C. nubeculosus* when combined with CO~2~, whereas octanal, nonanal, (*E*)-2-nonenal and 3-propylphenol acted as behavioral inhibitors. The behavioral effect of these volatiles was dose dependent.
The aim of this study was to determine the release rates of the compounds identified from cow urine and hair that either increase or reduce the attraction of *C. impunctatus* to traps baited with CO~2~-in the field. The experiments were conducted in Southern Sweden, an area where livestock are at risk to midge-borne diseases (Doréa et al. [@CR20]; Hultén et al. [@CR26]), and where biting midge populations feed voraciously on people and animals during the period between April to October (Ander et al. [@CR1]). We discuss our findings in relation to their potential use in future surveillance and control strategies targeting biting midges.
Methods and Materials {#Sec2}
=====================
Study Site {#FPar1}
----------
Field trapping was performed at Stockhultsgården, 14 km northwest of Markaryd, Sweden (N 56° 32,867′, E 13° 32.542′), from June to mid-July, the peak season for blood-feeding *C. impunctatus* in Southern Sweden. The field site, a meadow (approximately 2.6 ha), surrounded by a mixture of evergreen and deciduous trees, sustains a large population of *C. impunctatus*. As a well-established hunting ground, the area is inhabited by populations of elk, fallow deer, roe deer and wild boar, all potential hosts of biting midges (Blackwell et al. [@CR7]; Pettersson et al. [@CR36]; Viennet et al. [@CR54]).
Odour Compounds {#FPar2}
---------------
The compounds tested in this study have previously been identified in aerations of cattle hair and urine, and shown to elicit electrophysiological and behavioral responses in *C. nubeculosus* in the laboratory (Isberg et al. [@CR27]). Chemicals used were sourced from Sigma-Aldrich Chemie GmbH, Steinheim, Germany, except for 3-propylphenol which was obtained from Alfa Aesar GmbH Karlsruhe, Germany (Table [1](#Tab1){ref-type="table"}).Table 1Odor compounds and dispensers used in combination with carbon dioxide in field tests of trapping *Culicoides* biting midges and target release ratesCompoundPurity (%)OriginRelease rateapprox 0.1 mg h^−1^approx 1 mg h^−1^approx 10 mg h^−1^No. vialsHole dia (mm)No. vialsHole dia (mm)No.vialsHole dia (mm)Octanal98Cattle hair1426216Nonanal95Cattle hair18210616Decanal98Cattle hair18116916(*E*)-2-nonenal97Cattle hair161169162-Ethylhexanol99Cattle urine162109161-Octen-3-ol98Cattle hair14110516Phenol99Cattle urine161105open^a^4-Methylphenol99Cattle urine162106open^a^3-Propylphenol98Cattle urine1821010open^a^Dispensers were polyethylene vials (height 32.9 mm, diameter 22.8 mm) with a hole drilled in the lid^a^Lid removed from vial
Determination of Release Rates {#FPar3}
------------------------------
Test compounds (100 μl) were released via holes drilled in the lids of polyethylene vials (Kartell Labware, Noviglio, Italy; height 32.9 mm, diameter 22.8 mm, wall thickness 1.35 mm) at three different approximate release rates of 0.1 mg h^−1^, 1 mg h^−1^ and 10 mg h^−1^. The size of the holes and number of vials required for each combination of compound and release rate (Table [1](#Tab1){ref-type="table"}) were determined through preliminary experiments. Vials were weighed prior to and following 24 h of outdoor exposure, with weights recorded every hour for the first 6 h. This procedure was repeated twice, using two vials per compound per repetition, to obtain an average release rate for each compound to be used in trapping experiments (Table [2](#Tab2){ref-type="table"}). Maximum and minimum temperatures during the measurement period were 19 °C and 8 °C respectively, similar to those during the period of trapping experiments (16 °C and 9 °C respectively).Table 2Average release rates (± SEM) over 24 h of odor compounds from vials with lids with different sized holes (*N* = 4; 8 °C -- 19 °C)CompoundLid hole diameter (mm)Release rate (mg h^−1^)Octanal40.18 ± 0.0760.48 ± 0.27165.47 ± 0.38Nonanal80.12 ± 0.03100.50 ± 0.14161.62 ± 0.18Decanal80.08 ± 0.12161.12 ± 0.03(*E*)-2-nonenal60.18 ± 0.14161.15 ± 0.112-Ethylhexanol60.20 ± 0.09100.43 ± 0.17161.12 ± 0.171-Octen-3-ol40.18 ± 0.10101.05 ± 0.07162.05 ± 0.48Phenol60.15 ± 0.03100.93 ± 0.28open1.92 ± 0.344-Methylphenol60.10 ± 0.08100.60 ± 0.15open1.58 ± 0.383-Propylphenol80.15 ± 0.10100.48 ± 0.20open0.93 ± 0.20
Trapping Protocol {#FPar4}
-----------------
The field site was divided into five sub-sites, each separated by at least 100 m. To avoid cross contamination, only one compound was tested at each sub-site at any one time, and each compound was only tested at one sub-site. Traps used were Centers for Disease Control and Prevention (CDC) standard miniature light traps fitted with a CO~2~ delivery system (Model 1012-CO~2~; The John W. Hock Company, Gainesville, Florida). Carbon dioxide from a cylinder (Strandmöllen AB, Ljungby, Sweden) was released at 500 ml min^−1^ (Bhasin et al. [@CR5]; Harrup et al. [@CR24]) from both control and test traps. The vials containing test compounds were hung on the outside of the trap close to the CO~2~-release point of the test traps (Fig. [1](#Fig1){ref-type="fig"}).Fig. 1Placement of vials releasing odor compounds on the CDC light trap
A randomized design was used to study the individual odor compounds at different release rates. One control trap, releasing CO~2~ only, was placed with the three traps baited with the individual odor compounds released at different rates, together forming a test group. One test group was placed at one of the sub-sites in the form of a square, with traps 1.5 m from the ground, 3 m apart. This experimental set up was selected to counteract high temporal and spatial variation of biting midge densities that can increase or decrease dramatically over distances less than 50 m (Kirkeby et al. [@CR29]; personal observations). The positions of the traps were randomly placed within the test group before every test night, using a computerized randomization scheme (Minitab® 15 Statistical Software, Minitab Inc. UK).
Each trap was set to operate from 2 h before sunset until 1 h after sunset, for a total of seven nights, with vials replaced after every night. Insects attracted to each trap were drawn into a collection bottle containing a dilute detergent solution. Bottles were emptied daily, and the insects placed in 75% ethanol for storage and transportation. In the laboratory, the numbers of biting midges collected were recorded and identified to species level under a microscope using a morphological identification key (Delecolle [@CR19]). Larger samples, \>1000 biting midges, were subsampled according to the method of Vanark and Meiswinkel ([@CR51]) to obtain an estimate of the total number of biting midges in the sample.
Statistical Analysis {#FPar5}
--------------------
The goal of the statistical analyses was to identify release rates for each compound, which, when used in combination with CO~2~, attracted significantly higher or lower numbers of biting midges than the CO~2~-only control. A mixed modelling approach was used to control for variation in numbers of biting midges collected between days (Paterson and Lello [@CR35]), using the lme4 (Bates et al. [@CR3]) package in R ([@CR41]). Biting midge numbers were log-transformed prior to analysis to control for over-dispersion in the distribution of trap collections.
For each compound, the log-transformed number of biting midges was entered as the dependent variable in the model, with compound release rate entered as a four-level fixed factor (CO~2~-only control set as the reference level of the factor, 0.1 mg h^−1^, 1 mg h^−1^, 10 mg h^−1^). Day (experimental days 1--7) was entered as a random effect. To identify whether there was a significant overall effect of release rate on the number of biting midges collected, likelihood ratio tests were used to compare the residual deviance of models that included and excluded release rate as a factor. Where an overall effect of release rate was found, the significance of differences between number of biting midges caught at 0.1 mg h^−1^, 1 mg h^−1^ and 10 mg h^−1^ compared to the CO~2~-only control was assessed through the model coefficients associated with each release rate. Coefficients with absolute *t*-values greater than 2 were gauged to indicate significantly different (*P* \< 0.05) numbers of biting midges caught compared to the CO~2~ only control (Baayen et al. [@CR2]). Results were presented graphically by extracting the model predictions derived from the fixed effects only (release rate), which were subsequently back-transformed onto the original scale. Bootstrapped-confidence intervals (100 simulations) for the predictions were calculated using the boot package in R (Canty and Ripley [@CR12]; Davison and Hinkley [@CR18]), back transformed onto the original measurement scale.
Results {#Sec3}
=======
Total Number of *Culicoides* Collected and Species Identification {#FPar6}
-----------------------------------------------------------------
In total, 642,933 biting midges were collected in the control and odor-baited traps. Of these, 99.98% were identified as *C. impunctatus*, with *C. obsoletus senso lato* accounting for the remaining 0.024%.
Relative Attractiveness of Individual Compounds at Different Release Rates {#FPar7}
--------------------------------------------------------------------------
A significant effect of the release rate of nonanal on biting midges collected was detected (*χ* ^*2*^ = 8.40, *df* = 3, *P* \< 0.05; Fig. [2](#Fig2){ref-type="fig"}, middle left). However, none of catches with the different release rates differed significantly from the catch in the trap baited with CO~2~ only. This indicated that a significant difference existed between the number of biting midges collected by traps baited with nonanal released at 0.1 mg h^−1^ and 10 mg h^−1^ (Fig. [2](#Fig2){ref-type="fig"}). The release rate of decanal was found to have an impact on the numbers of biting midges collected (*χ* ^*2*^ = 22.1, *df* = 3, *P* \< 0.001; Fig. [2](#Fig2){ref-type="fig"}, bottom left), and traps baited with decanal released at the two lower rates collected a significantly higher number of biting midges than the CO~2~-only control trap. There was no significant difference in the number of biting midges collected in traps baited with decanal released at 10 mg h^−1^ or those baited with CO~2~ only.Fig. 2Predicted number of biting midges collected (±95 % confidence intervals. Fixed effects only) for nine cattle-derived odors, when released in combination with carbon dioxide (CO~2~). White bars represent CO~2~-only controls (*N* = 7), and grey bars CO~2~ plus test compound at three different release rates (*N* = 7). Asterisks indicate significant differences (*P* \< 0.05) from the CO~2~-only control in the mixed model
There was a significant overall effect of the octanal release rate on the numbers of biting midges collected (*χ* ^*2*^ = 15.7, *df* = 3, *P* \< 0.01; Fig. [2](#Fig2){ref-type="fig"}, top left). While there was no significant difference in numbers collected between traps baited with octanal released at 0.1 mg h^−1^ and the CO~2~-only control trap, traps baited with octanal released at 1 mg h^−1^ and 10 mg h^−1^ collected significantly fewer biting midges than the control trap. Similarly, there was an overall effect of the (*E*)-2-nonenal release rate on the numbers of biting midges collected (*χ* ^*2*^ = 8.8, *df* = 3, *P* \< 0.05; Fig. [2](#Fig2){ref-type="fig"}, top center), and traps baited with (*E*)-2-nonenal at the two highest release rates collected significantly fewer biting midges than the CO~2~-only control trap. However, there was no significant difference in numbers collected between traps baited with the lowest release rate of (*E*)-2-nonenal and the CO~2~ control trap.
The release rate of 2-ethylhexanol had no effect on the number of biting midges collected (*χ* ^*2*^ = 5.6, *df* = 3, *P* \> 0.05; Fig. [2](#Fig2){ref-type="fig"}, center). In contrast, the release rate of 1-octen-3-ol, had an overall impact on the number of biting midges collected (*χ* ^*2*^ = 16.3, *df* = 3, *P* \< 0.001; Fig. [2](#Fig2){ref-type="fig"}, bottom center). Traps baited with 1-octen-3-ol released at the two lower rates collected a significantly higher number of biting midges than the CO~2~ control trap, whereas there was no significant difference in numbers collected in the trap baited with 1-octen-3-ol released at 10 mg h^−1^ and the CO~2~-only trap.
The release rate of phenol was found to have an impact on the number of biting midges collected (*χ* ^*2*^ = 7.9, *df* = 3, *P* \< 0.05, Fig. [2](#Fig2){ref-type="fig"}, top right). Traps baited with phenol released at 0.1 mg h^−1^ and 1 mg h^−1^ collected more biting midges than the CO~2~-only control trap. There was no significant difference in numbers collected between traps releasing phenol at the highest release rate and the control trap. Similarly, the release rate of 4-methylphenol also had an impact on the number of biting midges collected (*χ* ^*2*^ = 16.4, *df* = 3, *P* \< 0.01, Fig. [2](#Fig2){ref-type="fig"}, middle right). Traps baited with 4-methylphenol released at the two lowest rates collected more biting midges than the CO~2~-only control trap. The number of biting midges collected also varied with release rate of 3-propylphenol (*χ* ^*2*^ = 12.2, *df* = 3, *P* \< 0.01; Fig. [2](#Fig2){ref-type="fig"}, bottom right), with traps baited with 3-propylphenol at all release rates tested collecting a significantly higher number of biting midges than the CO~2~-only control trap.
Discussion {#Sec4}
==========
Semiochemicals have a demonstrated potential for use in control strategies targeting biting midges (Carpenter et al. [@CR14]). Building on laboratory results on *C. nubeculosus*, this study showed that some cattle-associated kairomones can be used to enhance attraction of *C. impunctatus* to CO~2~-baited traps in the field. Moreover, some other kairomones reduced trap captures, indicating their potential use as repellents.
Low release rates of decanal, in combination with CO~2~, increased trap captures of *C. impunctatus* when compared to the control CO~2~ trap. This is in agreement with previous studies showing that host-derived aldehydes, either alone or in combination with other host volatiles, play an important role in the attraction of mosquitoes (Syed and Leal [@CR43]; Tchouassi et al. [@CR46]) and tsetse flies (Gikonyo et al. [@CR23]). Tchouassi et al. ([@CR46]) also showed that blends of host-derived aldehydes, in combination with CO~2~, were more effective in attracting mosquitoes compared to individual aldehydes. Similar results have been obtained from studies on herbivorous insects showing that behavioral responses to blends of host volatiles often exceed the responses to individual components (Pickett et al. [@CR37]; Webster et al. [@CR55]). An implication of this is that the volatiles may be perceived as non-host cues if detected by the insect individually, outside the context of the blend, but when combined together in a blend they may be perceived as an attractive host stimulus. This may be the case for octanal, nonanal and (*E*)*-*2-nonenal, which, if presented at a lower release rate or in a blend, could be perceived as a host-cue by biting midges. Alternatively, octanal and (*E*)*-*2-nonenal may act as host-derived repellents at higher doses (Jaleta et al. [@CR28]).
The results obtained for 1-octen-3-ol are consistent with those of previous field studies on *C. impunctatus*, *C. nubeculosus* and other biting midge species (Bhasin et al. [@CR5]; Blackwell et al. [@CR9]; Harrup et al. [@CR24]; Kline et al. [@CR30]; Ritchie et al. [@CR42]). 1-Octen-3-ol is a well-characterized mammalian kairomone (Pickett et al. [@CR37]), which is known to attract various hematophagous insects, including tsetse flies (Torr [@CR47]; Vale and Hall [@CR49]) and mosquitoes (Kline et al. [@CR30]; Takken and Kline [@CR44]). As was observed for decanal, *C. impunctatus* were more attracted to traps with a low release rate of racemic 1-octen-3-ol, as previously shown in both laboratory and field studies (Bhasin et al. [@CR4]; Blackwell et al. [@CR9]; Isberg et al. [@CR27]) of *C. impunctatus* and *C. nubeculosus*. While this study analyzed the effect of racemic 1-octen-3-ol on attraction of *C. impunctatus*, other studies on both biting midges and mosquitoes imply that it is the (*R*)*-*enantiomer that is important for attraction (Harrup et al. [@CR24]) and for some species even repellence (Pingxi et al. [@CR38]). Based on the results of Harrup et al. ([@CR24]), future field experiments on *C. impunctatus* should validate the effect of (*R*)*-*1-octen-3-ol seen in *C. nubeculosus* and *C. obsoletus.*
Addition of phenol, 4-methylphenol, 3-propylphenol and 2-ethylhexanol, identified in air entrainments of cattle urine (Isberg et al. [@CR27]), to traps baited with CO~2~ increased trap captures of *C. impunctatus*. The results obtained for phenol, as well as 4-methylphenol, agree with those observed for *C. nubeculosus* in the laboratory (Isberg et al. [@CR27]). Phenol, 4-methylphenol and 3-propylphenol, in combination with other host volatiles, have previously been shown to attract *C. impunctatus* (Bhasin et al. [@CR5]) as well as other biting midge species (Cilek et al. [@CR17]; Venter et al. [@CR52]) when compared to an unbaited control traps. Phenolic compounds found in ox urine are also known attractants for tsetse flies (Bursell et al. [@CR11]; Vale et al. [@CR50]) and the zoophilic mosquito *Anopheles quadriannulatus* (Takken and Knols [@CR45]). Unlike tsetse flies that are attracted to blends of phenolic compounds in urine to a level equal to or greater than those with natural urine (Bursell et al. [@CR11]; Torr et al. [@CR48]; Vale et al. [@CR50]), the results from this study suggest that biting midges are attracted to individual phenolic compounds at a similar level to that of natural urine; phenol (0.1 mg h^−1^), 4-methylphenol (0.16--1.41 mg h^−1^) and 3-propylphenol (0.023--0.18 mg h^−1^), and in the case of 3-propylphenol also higher rates than the natural release. In the field experiments presented here, 2-ethylhexanol, when released in combination with CO~2~, did not collect significantly more biting midges than the CO~2~ only control trap, although a larger capture was observed when 2-ethylhexanol was released at 0.1 mg h^−1^. This suggests that lower release rates of 2-ethylhexanol should be tested in future field experiments, an argument also supported by behavioral experiments in the laboratory with *C. nubeculosus* (Isberg et al. [@CR27]).
The attraction of *C. impunctatus* to traps baited with cattle-derived kairomones provides a more diversified set of control and surveillance tools than that currently available (Carpenter et al. [@CR14]). While semiochemical-based trapping of biting midges is likely to reflect host-seeking populations more accurately than other trapping protocols, the current lures, predominantly 1-octen-3-ol and CO~2~, are not optimal (Carpenter et al. [@CR14]; Carpenter et al. [@CR15]; Gerry et al. [@CR21]; Harrup et al. [@CR24]; Viennet et al. [@CR53]). Future studies will have to assess whether the compounds originally identified to modulate the behavior of *C. nubeculosus* in the laboratory, and then of *C. impunctatus* in the field, also attract other *Culicoides* species. Further optimization of attractive lure(s) could involve analyzing the effect of blends, including host-derived aldehydes or phenolic compounds, on biting midge behavior. The findings that some host-derived volatiles can inhibit the host seeking behavior of *Culicoides* biting midges (Isberg et al. [@CR27]; this study) merit further investigation. These repellents might be an innovative way to disrupt the host-seeking behavior of biting midges, similar to what has been shown for tsetse flies (Gikonyo et al. [@CR22]) and malaria mosquitoes (Jaleta et al. [@CR28]).
This project was financed by the Swedish Research Council (Formas) and supported by the Linnaeus Grant IC-E3 (Formas, Sweden). DPB was supported by an EU intra-European fellowship for career development.
| {
"pile_set_name": "PubMed Central"
} |
One of the vital hallmarks of cancer which distinguishes it from benign tumours is its capability for tissue invasion and distant metastasis ([@bib26]). Focal adhesion kinase (FAK) is a non-receptor protein tyrosine kinase identified as a critical signalling molecule mediating host--tumour interactions that affect cell adhesion, invasion, angiogenesis and metastasis via its cross-linked processes with Src-, integrin- and growth factor receptor signalling pathways ([@bib27]; [@bib33]). FAK is overexpressed in various malignancies, including lung cancer, wherein associated FAK gene amplification portends a worse prognosis ([@bib2]; [@bib7]). Indeed, FAK activation has been implicated in conferring a more invasive phenotype to malignancies through its effects on the epithelial-to-mesenchymal transition process by promoting E-cadherin delocalisation and upregulation of caveolin-1 ([@bib3]; [@bib11]).
Y15 is a small molecule identified through *in silico* screening that selectively targets the Y397 autophosphorylation site of FAK ([@bib14]). This is the same interaction site between activated FAK and Src, as well as the binding site for other signalling proteins ([@bib31]; [@bib9]). Y15 has shown activity, either alone or in combination with chemotherapy, both *in vitro* and *in vivo*, against breast, pancreatic, neuroblastoma and colon cancer tumour growth ([@bib14]; [@bib20]; [@bib5]; [@bib19], [@bib18]).
The FAK-Src signalling network appears to be deregulated in high-grade lung cancers. Phosphoproteomic analysis comparing KRAS-mutant/LKB1 wild-type lung tumours with KRAS-mutant/LKB1-deficient lung tumours revealed hyperphosphorylation of FAK and Src which promoted cell migration in the latter ([@bib8]). This may provide mechanistic explanation for the observation that inactivation of the tumour suppressor LKB1 results in greater propensity for metastasis compared with loss of other tumour suppressors in a KRAS-driven murine lung cancer model ([@bib10]). We thus sought to characterise the effects of FAK inhibition using Y15 and other FAK inhibitors in various lung cancer cell lines with or without RAS mutations.
Materials and methods
=====================
Cell lines
----------
KRAS-mutant (H157, H358, H727, A549), NRAS-mutant (H1299) and EGFR-mutant (H1650, H1975) cell lines were obtained from the American Type Culture Collection. Cell lines were cultured in RPMI-1640 (Life Technologies, Grand Island, NY, USA) supplemented with 10% FBS (Sigma-Aldrich, St Louis, MO, USA) and penicillin--streptomycin and incubated at 37 °C in a fully humidified atmosphere containing 5% CO~2~. All cell lines were used at low passage in our laboratory, and were authenticated using viability, morphology and growth curve analysis on a regular basis, and tested negative for Mycoplasma.
Compounds
---------
Y15 and the ATP-competitive FAK inhibitor PF-573228 were purchased from Sigma-Aldrich. Other ATP-competitive FAK inhibitors PF-562271 and NVP-TAE226 (TAE-226), as well as Bcl-2 inhibitors ABT263, GX15-070 and ABT-737 were obtained from Selleckchem (Houston, TX, USA). FAK inhibitor C4 ([@bib25]) was obtained from the laboratory of Dr Elena Kurenova and was used as control in the MTS assays to screen for the effect of Y15 on lung cancer cell viability.
Antibodies
----------
The antibodies of p-FAK (Y397), FAK, p-Akt (S473), Akt, p-Erk1/2 (T202/Y204), Erk, p-Src (Y416), Src, cleaved caspase-3, cleaved poly (ADP-ribose) polymerase (PARP), p-STAT3 (Y705), p-P70S6K (T389), p-mTOR (S2448), Bcl-2, Bcl-xL and Mcl-1 were obtained from Cell Signaling Technology (Danvers, MA, USA). p-JNK (T183) and JNK1 antibodies were from Santa Cruz Biotechnology (Dallas, TX, USA). *β*-Actin antibody was purchased from Sigma-Aldrich.
Clonogenicity assays
--------------------
In all, 5 × 10^5^ cells were seeded into 6-well plates and treated with escalating dosages of Y15 (or DMSO) and incubated at 37 °C for 14--20 days. Media were changed every 3--4 days. Colonies were then visualised and counted after staining with crystal violet.
Cell-viability assays
---------------------
Initial screening of Y15 effect on lung cancer cell line viability was performed using trypan blue exclusion assay as previously described ([@bib41]) using C4 as comparator agent. MTS assay was used for all subsequent experiments. Briefly, cells of 1 × 10^4^ were seeded in 96-well plates and treated with concurrent administration of Y15 with each of the drugs for 48 or 96 h. Then 20 *μ*l per well of CellTiter 96 AQueous One Solution Reagent (Promega, Madison, WI, USA) was added. After 1 h at 37 °C in a humidified, 5% CO2 atmosphere, the absorbance at 490 nm was recorded using a microplate reader.
Western blot analysis
---------------------
For immunoblot analysis, the cells were treated with the indicated agents and then collected in cell lysis buffer (Cell Signaling Technology). Total protein was quantified using Coomassie protein assay reagent (Bio-Rad, Hercules, CA, USA). An equal amount of protein (60 *μ*g) was separated by SDS-PAGE and electrotransferred onto nitrocellulose membrane. The primary antibodies were then tested. *β*-Actin was used as endogenous control for equal loading. Immunocomplexes were visualised with enhanced chemiluminescence detection kits (Pierce, Rockford, IL, USA).
RNA interference studies
------------------------
RNA interference was based on pGreenPuro system (System Bioscience, Mountain View, CA, USA) expressing small-hairpin RNA (shRNA). pGreen-Akt1, pGreen-Bcl-2, pGreen-Bcl-xL and pGreenPuro-vec constructs, encoding shRNA for Akt1 (shAkt1), Bcl-2 (shBcl-2), Bcl-xL (shBcl-xL) or a negative control (vector), respectively, were prepared by inserting the target sequence for human Akt1 (5′-ATGACTTCCTTCTTGAGGATC-3′) or Bcl-2 (5′-CCCTGATTGTGTATATTCA-3′) or Bcl-xL (5′-CCCTTGCAGCTAGTTTTCT-3′) into pGreenPuro. 293TN cell was stably transfected with the constructs and three packaging plasmids using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) to package lentivirus; and then lung cancer cells were infected with lentivirus with multiplicity of infection of 5. Clones with stable downregulated Akt1, Bcl-2 or Bcl-xL expression were selected with puromycin (1 *μ*g ml^−1^).
Small-interfering RNA (siRNA) targeting JNK (MAPK8 Silencer Select siRNA S11154) were purchased from Life Technologies. Transient transfection of siRNAs was carried out using Lipofectamine 2000 (Life Technologies) according to the manufacturer\'s instructions. Briefly, cells were seeded in 24-well plate. On the second day, cells were transfected with 20 pMol siRNA per well with lipofectamine 2000 (Invitrogen), and media were changed 4 h after transfection. Twenty-four hours later, cells were then treated with Y15 for cell-viability analysis.
*In vivo* xenograft experiments
-------------------------------
All *in vivo* experimental protocols were approved by the Institutional Animal Care and Use Committee of Roswell Park Cancer Institute (RPCI; Buffalo, NY, USA). Female SCID mice, 6--8-week old, were used for the experiments. Lung cancer cells (5 × 10^6^) were injected s.c. into the flank of SCID mice (RPCI). Tumours were monitored until they reached a mean tumour volume of 100 or 250 mm^3^ before starting Y15 dosing. Mice were assigned randomly to different groups (five mice per treatment group).Y15 was administered by oral gavage once daily at respective doses shown in the accompanying figures (see the 'Results\' section). Tumour volume was measured in two dimensions (length and width) twice-weekly using Ultra Cal-IV calipers (Fred V. Fowler Company, Inc., Newton, MA, USA) and was analysed using studylog software (Studylog Systems, San Francisco, CA, USA). Tumour volume (mm^3^)=(length × width^2^)/2. Mouse body weights were also recorded twice-weekly and the mice were observed daily. Mice with tumour volumes ⩾2000 mm^3^ or with losses in body weight ⩾20% from their initial body weight were promptly killed per Institutional Animal Care and Use Committee guidelines.
Immunohistochemistry
--------------------
Immunohistochemistry was performed in Pathology department of RPCI as previously described ([@bib32]). Briefly, tumour tissue was fixed overnight in 10% neutral-buffered formalin at room temperature, transferred to 70% ethanol and processed for paraffin embedding using a Thermo Electron Excelsior tissue processor (Thermo Fisher Scientific, Pittsburgh, PA, USA). Paraffin blocks were sectioned to 4-mm thickness and placed on positively charged glass slides. Tissues were stained using a Discovery automated slide machine (Ventana Medical Systems, Tucson, AZ, USA). Sections were counter-stained with hematoxylin to enhance visualisation of tissue morphology. After treatment in terminating buffer (300 m[M]{.smallcaps} sodium chloride and 30 m[M]{.smallcaps} sodium citrate), the sections were incubated in streptavidin--peroxidase complex for 30 min and then developed with diaminobenzidine-tetra-hydrochloride for 1--5 min as a substrate.
Statistical analysis
--------------------
The reported values represent the means±s.d. for at least three independent experiments performed in triplicate. To determine the significance between the tested groups, Student\'s test was used, with *P*\<0.05 considered as significant. Median dose--effect analysis was performed using CalcuSyn software version 2 (Biosoft, Great Shelford, Cambridge, UK) to determine additive, synergistic or antagonistic interactions of drug combinations.
Results
=======
Y15 decreases lung cancer cell viability and clonogenicity *in vitro* in a dose-dependent manner
------------------------------------------------------------------------------------------------
To characterise the effect of FAK inhibition using Y15 in various lung cancer cell lines, the basal expression levels of Y397-pFAK and total FAK in several lung cancer cell lines were analysed ([Figure 1A](#fig1){ref-type="fig"}). Levels of Y397-pFAK and FAK were variable across cell lines. We then screened for the efficacy of Y15 against five cell lines with 3-day exposure to escalating doses of Y15. Results showed decreased cell viability by MTS assay, whereas the control agent C4, a FAK scaffold inhibitor which disrupts FAK-VEGFR3 signalling and is not anticipated to be cytotoxic in lung cancer cell lines, had virtually no effect ([Figure 1B](#fig1){ref-type="fig"}). MTS assay was also performed for the ATP-competitive small molecule FAK inhibitors PF-562271, PF-573228 and TAE-226. [Table 1](#tbl1){ref-type="table"} demonstrates comparative IC50 values as determined by MTS assay, showing that Y15 is more potent compared with the most selective FAK inhibitor PF-573228, with comparable to slightly more potent activity compared with PF-573228 and TAE-226 ([Table 1](#tbl1){ref-type="table"}). We also treated lung cancer cell lines for 72 h and determined IC50 values for Y15 by clonogenic assay ([Figure 1C](#fig1){ref-type="fig"}). Y15 decreased clonogenicity in a dose-dependent manner in multiple cell lines regardless of RAS mutation status ([Figure 1C](#fig1){ref-type="fig"}).
Y15 effectively inhibits the growth of lung cancer xenografts *in vivo*
-----------------------------------------------------------------------
To test the efficacy of Y15 on RAS-mutant lung cancer growth *in vivo*, we implanted A549 cells subcutaneously into mice. As shown in [Figure 2A](#fig2){ref-type="fig"}, Y15 significantly decreased tumour growth. The same effect was observed with H1299 xenograft ([Figure 2B](#fig2){ref-type="fig"}). Immunohistochemical staining demonstrated that Y15 decreased Y397 FAK phosphorylation and triggered activation of caspase-3 in the harvested A549 tumour xenografts compared with the PBS-treated group ([Figure 2C](#fig2){ref-type="fig"}). We observed similar growth inhibition in EGFR-mutant H1650 and H1975 xenografts ([Figure 2D and E](#fig2){ref-type="fig"}). There was no significant difference in body weights (data not shown) and side effects between the vehicle and Y15-treated animals even when the dosages evaluated. Thus, Y15 can inhibit the growth of lung cancers with oncogenically driven MAPK pathway activation through either RAS or EGFR mutation.
Y15 inhibits Y397 FAK autophosphorylation and downregulates Bcl-2, Bcl-xL and Mcl-1 in a time-dependent manner in non-small cell lung cancer cell lines
-------------------------------------------------------------------------------------------------------------------------------------------------------
To determine the time-dependent effect of Y15 on FAK autophosphorylation, we performed Western blotting on cell lysates at different time points from 0.5 through 72 h exposure to Y15 at 5 *μ*M concentration in the H1299 cells (IC50∼4 *μ*M). [Figure 3A](#fig3){ref-type="fig"} shows time-dependent reduction in Y397-pFAK level. We also detected immediate and sustained reduction of phosphorylation of downstream substrate of FAK: p-Src (Y416). There was decreased phosphorylation of Akt, STAT3, P70S6K, GSK3*β* and mTOR at 72 h as well. There was also evidence of pro-apoptotic effects with decrease in Bcl-2, Bcl-xL and Mcl-1 levels. Intriguingly, phosphorylation of JNK was significantly increased until the 24-h time point, wherein cleavage of PARP and caspase-3 began to increase. However, Erk1/2 (T202/Y204) phosphorylation was not affected. Indeed, reduction of ERK phosphorylation with Y15 treatment was cell line-dependent (data not shown). Similar results were observed in additional cell lines ([Figure 3B](#fig3){ref-type="fig"}). Activation of JNK was induced upon treatment of Y15 until 8-h time point, whereas Bcl-2, Bcl-xL and Mcl-1 were decreased with increasing drug exposure time course. In summary, Y15 caused time-dependent decrease of Y397-pFAK, Bcl-2, Bcl-xL and Mcl-1 across multiple cell lines.
Y15-induced downregulation of Bcl-2 family members is independent of caspase activation
---------------------------------------------------------------------------------------
Given that caspase activation induced protein degradation during cell apoptosis, we evaluated whether caspase was involved in the downregulation of Bcl-2 family members. H1299 cells treated with 5 *μ*[M]{.smallcaps} Y15 were cultured in the presence or absence of 20 *μ*[M]{.smallcaps} pan-caspase inhibitor, z-VAD-fmk, for 8--48 h. As shown in [Figure 3C](#fig3){ref-type="fig"}, reduction in both Bcl-2 and Bcl-xL was evident from 8 h onwards under Y15 treatment alone as expected. However, no significant difference in this effect was found with the addition of caspase inhibitor z-VAD. Similar result was found in Mcl-1 from 24--48 h treatment. Taken together, the above results showed that downregulation of Bcl-2 family members accompanying Y15-induced apoptosis was independent of caspase activation.
Activation of JNK contributes to Y15-induced downregulation of Bcl-2 family members
-----------------------------------------------------------------------------------
It was reported that JNK is involved in downregulation of Bcl-2 family members ([@bib38]; [@bib21]; [@bib4]). To elucidate the effects of JNK activation on induction of apoptosis, knockdown of JNK by siRNA (si-JNK) was used in selected Y15-sensitive cells (H1299, H1650 and H358). As shown in [Figure 4A](#fig4){ref-type="fig"}, in the presence of 5 *μ*[M]{.smallcaps} Y15 treatment alone, p-JNK (T183) was significantly activated, whereas both p-FAK (Y397) and p-Akt (S473) were inhibited in the tested cell lines. In contrast with the siRNA-negative control, the expression of JNK was totally abrogated by siRNA in H1299, H1650 and H358 si-JNK cell lines. Furthermore, the previously observed degradation of Mcl-1, Bcl-xL and Bcl-2 upon treatment with Y15 alone was substantially attenuated upon knockdown of JNK ([Figure 4A](#fig4){ref-type="fig"}). Notably, the decrease of Mcl-1 and Bcl-2 in Y15-treated si-JNK cells was minimal to absent, suggesting that JNK activation is the process upstream of Mcl-1 and Bcl-2 regulation, and exhibits a functional role in the attenuation of apoptosis in si-JNK cells. JNK downregulation by siRNA protected cells from Y15-induced cell death in H1299, H358 and H1650 cells ([Figure 4B](#fig4){ref-type="fig"}; *P*\<0.001, 0.00368, \<0.001, respectively. Raw data is provided in the [Supplementary Materials](#sup1){ref-type="supplementary-material"}), supporting the hypothesis that activation of JNK contributes to Y15-induced apoptosis. Together, these results demonstrate that JNK activation has an important role in Y15-induced downregulation of Bcl-2 family members.
Abrogation of anti-apoptotic Bcl-2 family proteins synergizes with inhibition of FAK in lung cancer
---------------------------------------------------------------------------------------------------
As there was decreased Akt phosphorylation and downregulation of both Bcl-2 and Bcl-xL upon treatment with Y15 over time, we hypothesised that targeting these proteins may modulate the efficacy of Y15. To determine the role of these proteins, we knocked down Bcl-2, Bcl-xL or Akt1 by shRNA in H1299 cells. Stable clones propagated after puromycin selection (1 *μ*g ml^−1^) were then treated with Y15 and the results were compared with parental H1299 cell lines using trypan blue exclusion method. [Figure 5A](#fig5){ref-type="fig"} shows that knockdown of either Bcl-2 or Bcl-xL significantly decreased cell viability upon exposure to Y15, whereas no additional effect was observed with knockdown of Akt1. [Figure 5B](#fig5){ref-type="fig"} shows the expression of Akt1, Bcl-2, Bcl-xL in infected clones *vs* control cells that express the vector alone. In contrast to treatment control cells, there was minimal or modest decrease of Mcl-1 observed in shAkt1 clones upon treatment with Y15. In comparison, Mcl-1 levels were markedly reduced in shBcl-xL and shBcl-2 clones treated with Y15.
To investigate the signalling effects with pharmacologic inhibition, the effects of the combination of Y15 and the Bcl-2/Bcl-xL inhibitor ABT263 are illustrated in [Figure 5C--E](#fig5){ref-type="fig"}. The combination of Y15 and ABT263 decreased Mcl-1 levels to a greater degree than either Y15 or ABT263 alone. This combination also abrogated Akt (S473) and STAT3 (Y705) phosphorylation to a greater degree than that can be achieved by Y15 alone ([Figure 5C](#fig5){ref-type="fig"}). Consistent with the signalling effects, cell-viability tests showed that either Y15 or ABT263 alone was minimally toxic to H1299 cells. However, exposure of cells to the combination markedly increased cell death (i.e., \>72% [Figure 5D](#fig5){ref-type="fig"}). Median dose--effect analysis yielded combination index values considerably\<1.0, corresponding to a highly synergistic interaction ([Figure 5E](#fig5){ref-type="fig"}).
MTS assay was thus subsequently performed to evaluate the effect of Y15 or other FAK inhibitors in combination with various Bcl-2 inhibitors across different lung cancer cell lines. [Supplementary Table S1](#sup1){ref-type="supplementary-material"} summarises the results. Synergism is demonstrated with the combination of FAK inhibitors with various Bcl-2 inhibitors across multiple lung cancer cell lines.
Discussion
==========
Metastatic non-small cell lung cancer remains a difficult-to-treat disease, and efforts to improve patient survival have yielded moderate success so far even with genotypically selected targeted therapies. Emerging evidence has implicated FAK to have a key signalling role that induces cancer cell proliferation, motility, survival, invasion and metastasis ([@bib13]). In this study, we demonstrated that treatment with Y15 significantly decreased viability and clonogenicity of several lung cancer cell lines. Y15 is a non-receptor tyrosine kinase FAK inhibitor, which effectively decreased Y397 FAK phosphorylation in a time- and dose-dependent manner, and also decreased phosphorylation of downstream FAK signalling players, such as Src, Akt and STAT3. Downregulation of Bcl-2 family members (Bcl-2, Bcl-xL and Mcl-1) was associated with Y15-induced apoptosis. This in turn requires activation of JNK, with subsequent Mcl-1 and Bcl-2 downregulation. Y15-induced downregulation of the anti-apoptotic Bcl-2 is an important characteristic in key contradistinction to other FAK inhibitors where there is no significant effect on Bcl-2 or Bcl-xL levels ([@bib39]). Although Mcl-1 downregulation was described with another FAK inhibitor ([@bib39]), the effect was transient, whereas this is sustained with exposure to Y15.
This is the first report to demonstrate the efficacy of FAK inhibition in H1975, an EGFR-mutant lung cancer cell line harbouring the T790M mutation resistant to first- and second-generation EGFR tyrosine kinase inhibitors (TKIs). This is an important finding as Y15 may demonstrate activity regardless of the EGFR mutation type, particularly as different EGFR mutations have varying sensitivity to currently used EGFR TKIs and drug resistance continues to be a significant clinical problem despite the development of third-generation TKIs. Moreover, we also demonstrated the efficacy of FAK inhibition in RAS-driven lung cancer, supporting observations reported previously that FAK inhibitors exert potent antitumour effects in mutant KRAS non-small cell lung cancer in association with INK4A/ARF deficiency ([@bib24]).
Recent reports demonstrated that the anti-apoptotic PI3K/Akt cascade promotes resistance to various anti-cancer therapies, including EGFR TKIs ([@bib36]; [@bib23]; [@bib22]).The Bcl-2 family also represents a critical group of molecules involved directly in the regulation of cell apoptosis ([@bib30]; [@bib1]; [@bib29]; [@bib16]). Indeed, the lack of efficacy of combined PI3K/Akt and MEK pathway inhibition in KRAS-mutant lung cancers can be attributed to the inability to induce apoptosis, as inhibition of Bcl-xL restores cytotoxicity and the apoptotic response ([@bib17]). We have demonstrated that Y15 induces apoptosis. Our results showed that Bcl-2, Bcl-xL and Mcl-1 downregulation contributed significantly to Y15-induced apoptosis. In addition, combination therapy with Bcl-2/Bcl-xL inhibitor, such as ABT263, potentiates Y15-mediated apoptosis ([Figure 5D and E](#fig5){ref-type="fig"}). Similar to what has been recently demonstrated in ovarian clear cell carcinoma cell lines ([@bib39]), we also demonstrated that pharmacological inhibition of Bcl-2/Bcl-xL in combination with various FAK inhibitors including Y15 has synergistic activity in lung cancer cell lines. One potential mechanism for the synergistic effect between Bcl-2 pathway inhibition with kinase inhibitors had been demonstrated previously by multiple groups of investigators ([@bib12]; [@bib15]; [@bib28]; [@bib35]; [@bib34]; [@bib40]). These studies demonstrate that upregulation of BIM, a pro-apoptotic member of the Bcl-2 family, improves that apoptotic effect induced by various kinase inhibitors across different malignancies, including MAPK pathway-activated cancers, such as RAS-mutant or EGFR-mutant in lung cancer. Conversely, polymorphic variants that lead to BIM isoforms lacking the pro-apoptotic BH3 domain may underlie innate primary resistance to therapy. Drugs such as ABT263 sequester Bcl-2 and Bcl-xL to release BIM, thereby enhancing the efficacy of kinase inhibitors. Moreover, paxillin, an adaptor protein that is phosphorylated by FAK and Src, is shown to confer resistance to EGFR TKIs in EGFR-mutant lung cancers by reducing BIM and increasing Mcl-1 expression ([@bib37]). Finally, inactivation of STAT3 by Y15 contributes to reduction in Bcl-xL expression, thereby enhancing the apoptotic potential of the combination regimen ([@bib42]; [@bib40]).
We conducted experiments to describe another mechanism of the synergism observed between FAK inhibition and Bcl-2 pathway inhibition. JNK is a member of the MAPK family and is activated by a variety of environmental stresses, inflammatory cytokines, growth factors and GPCR agonists ([@bib6]). Recent studies demonstrated that JNK is involved in downregulation of Bcl-2 family members ([@bib38]; [@bib21]; [@bib4]). In our study, JNK activation has a role in Y15-induced downregulation of Mcl-1 and Bcl-2. Knockdown of JNK by shRNA abrogated the downregulation of Mcl-1 and Bcl-2 by Y15 and attenuated Y15-induced cell lethality, indicating that JNK activation is the upstream process of Mcl-1 and Bcl-2 regulation. Altering JNK function or activity may thus affect the efficacy of Y15, or potentially, as well as on the combination of FAK and Bcl-2/Bcl-xL inhibitors.
In summary, this report for the first time demonstrates the effect of FAK inhibition alone or in combination with depletion of Bcl-2 pathway in oncogenically driven, MAPK-activated lung cancers through either RAS or EGFR mutation. FAK signalling is a novel therapeutic target worth exploring in the treatment of lung cancer, and the data shown here provide the rationale for further clinical development.
Major funding for this project was received from RPCI Alliance Foundation. This work was also made possible in part by National Institute of Health Grant, CA65910 and P30 Grant CA016056 from NCI to Roswell Park Cancer Institute. We also thank the Support in part by the Fundamental Research Funds for the Central Universities, Shaanxi Normal University (GK201603062). We thank Mike Yemma and Baotran Ho for their assistance with the harvest of A549 xenograft experiments for histopathological analysis.
[Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies this paper on British Journal of Cancer website (http://www.nature.com/bjc)
This work is published under the standard license to publish agreement. After 12 months the work will become freely available and the license terms will switch to a Creative Commons Attribution-NonCommercial-Share Alike 4.0 Unported License.
The authors declare no conflict of interest.
Supplementary Material {#sup1}
======================
######
Click here for additional data file.
######
Click here for additional data file.
{#fig1}
{#fig2}
![**Y15 inhibits Y397 FAK autophosphorylation and downregulates Bcl-2 family members.**(**A**) Y15 decreased Y397-pFAK and downstream p-Src, p-Akt targets in a time-dependent manner. H1299 lung cancer cell line was treated with 5 *μ*[M]{.smallcaps} of Y15 for 0--72 h and western blotting was performed with Y397-pFAK, FAK, p-Src, Src, p-Akt, Akt and other targets. Y15 increased cleaved PARP and caspase-3 starting from 24 h. (**B**) Y15 performed similar effects in H727 and H1650 in a time-dependent manner. (**C**) Downregulation of Bcl-2 family members was independent of caspase activation. Twenty *μ*[M]{.smallcaps} of broad-spectrum caspase inhibitor, z-VAD-fmk, was treated with or without 5 *μ*[M]{.smallcaps} Y15 at different time points after 8--48 h in the H1299 cells. Western blotting was performed to evaluate the effect on Bcl-2 family members.](bjc2016190f3){#fig3}
![**Knockdown of JNK by siRNA significantly attenuates Y15-induced cytotoxicity, and subsequently abrogates Mcl-1 and Bcl-2 downregulation induced by Y15.**H1299, H1650 or H358 cells were transfected with siRNA targeting JNK and were treated with Y15 for 48 h, after which whole-cell lysates were prepared and subjected to western blot analysis (**A**). The extent of apoptosis was determined as described in the 'Materials and Methods\' section (**B**). Twenty-four hours after siRNA transfection, cells were treated with Y15 (2.5 *μ*[M]{.smallcaps} for H1299, and 5 *μ*[M]{.smallcaps} for H358 and H1650) for 24 h. Then cells were collected and live cells were quantified with trypan blue exclusion methods. Results of Y15-treated cells were presented as % relative to values of vehicle-treated cells (defined as 100%). Data represents means±s.d. of three independent experiments (\**P*\<0.05, compared with scrambled siRNA control).](bjc2016190f4){#fig4}
![**Abrogation of anti-apoptotic Bcl-2 family members synergizes the antitumour efficacy of Y15 against lung cancer cells *in vitro*.**(**A**) H1299 cells with empty vector and their respective shBcl-2, shBcl-xL or shAkt1 clones were treated with Y15 for 72 h, trypan blue assay was performed to evaluate cell viability. (**B**) Western blots showing differences in the expression of Bcl-2, Bcl-xL and Akt1 in control H1299 and respective shRNA clone cells, with or without Y15. (**C**) Western blot showing the effect on downstream signalling proteins with the combination of Y15 at 5 *μ*[M]{.smallcaps} and the Bcl-2/Bcl-xL inhibitor ABT263 at 2.5 *μ*[M]{.smallcaps} in H1299 cells after 24 h treatment. The combination of Y15 and ABT263 decreased Y397-pFAK, p-Src and p-Akt more than each agent alone. (**D**) ABT263 potentiates Y15-mediated apoptosis. The cell growth inhibition was determined as described in the 'Materials and Methods\' section. Data represent means±s.d. of three independent experiments. (**E**) Median dose--effect analysis was performed to calculate CI using CalcuSyn software (Biosoft) to determine interactions of Y15 combined with ABT263.](bjc2016190f5){#fig5}
###### The effect of Y15 and other FAK inhibitors on lung cancer cell line viability (MTS assay)
**Cell line** **H157** **H358** **H727** **H1299** **A549** **H1650** **H1975**
------------------------------- ---------- ---------- ---------- ----------- ---------- ----------- -----------
**IC50 (*μ*[M]{.smallcaps})**
TAE226 7.46 3.18 0.30 2.98 2.59 2.30 2.88
PF-562271 5.76 6.38 2.26 3.94 5.41 5.31 5.47
PF-573228 11.39 27.49 2.80 9.18 9.08 20.70 13.20
Y15 2.10 2.72 1.30 3.88 3.49 1.90 1.56
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Myocarditis can present with a wide range of symptoms, ranging from mild dyspnea to chest pain, cardiogenic shock, and fatal arrhythmia. The main cause of myocarditis is current or recent viral infection \[[@B1]\]. Enteroviruses, specifically Coxsackievirus (CV) group B serotypes, have traditionally been perceived as the predominant viral cause \[[@B2]\], although adenoviruses, parvovirus B19, and human herpesvirus 6 can also cause myocarditis \[[@B3]-[@B5]\].
The pathophysiological progression of viral myocarditis includes three distinct phases \[[@B3]\]. The first phase is characterized by a nonspecific innate immune response, causing virus-mediated cell lysis and the indirect destruction of cardiomyocytes \[[@B3],[@B6]\]. During the second phase, a virus-specific immune response, including CD8^+^ lymphocytes, acts to eliminate the causative viruses, leading to heart failure with the destruction of the infected cardiomyocytes \[[@B3]\]. In the third phase, commonly a few weeks after infection, the destroyed cardiomyocytes are replaced by diffuse fibrosis and progressive biventricular dilatation, resulting in cardiac failure \[[@B1],[@B3]\]. Approximately 50% of patients with viral myocarditis develop chronic myocarditis, and 21% develop dilated cardiomyopathy (DCM) \[[@B7]\]. A longitudinal study reported that the immune clearance of viruses during or after the acute phase of myocarditis correlates with improvements in the left ventricular ejection fraction (LVEF) \[[@B8]\]. In chronic myocarditis or DCM, a persistent viral presence on endomyocardial biopsy (EMB) specimens is associated with an increased mortality rate \[[@B9]\]. However, viral genomic RNA and capsid protein are detectable in EMB specimens in no more than 35% and 10% of cases, respectively \[[@B6]\]. Therefore, the relationship between persistent viral infection and progression from acute myocarditis to chronic irreversible cardiomyopathy is unclear.
We present here a rare case of CVA4 infection causing acute myocarditis with concomitant pancreatitis and liver dysfunction. The patient's antibody titers against CVA4 were significantly elevated during the recovery period, and a pathological examination of an EMB specimen showed interstitial infiltration with CD3-positive lymphocytes, despite a normal LVEF on left ventriculography, confirming acute myocarditis. His plasma brain natriuretic peptide (BNP) and serum troponin I (TnI) levels remained elevated. Five weeks after admission, the patient underwent radical low anterior resection for advanced rectal cancer, found by chance. Interestingly, his plasma BNP and serum TnI levels returned to normal after surgery. The patient's clinical course suggests that the immune modulation associated with his rectal cancer and surgery favorably affected his acute myocarditis, preventing its progression to chronic myocarditis or DCM.
Case report
===========
A 63-year-old man was admitted to our hospital with a three-day history of dyspnea and fatigue, which had gradually increased until he experienced dyspnea and fatigue at rest. He had taken amlodipine (5 mg/day) for hypertension for the preceding five years, but had no other remarkable past history or family history. He was a nonsmoker and did not consume alcohol. Two weeks previously, he had presented with flu-like symptoms, including fever, sore throat, cough, and diarrhea, which had completely resolved within a week.
On admission, his height was 168.3 cm, weight 65.0 kg, and temperature 35.8°C. He was hypotensive, with a blood pressure of 72/52 mmHg, but his heart rate was not elevated (61 beats/min). A physical examination revealed cyanosis of the lips, distended external jugular veins, pretibial edema in both legs, coarse crackles over the lower bilateral lung fields, and mild enlargement of the liver. He was obviously short of breath in room air, with blood O~2~ saturation of 93%, partial O~2~ pressure of 61.7 mmHg, and partial CO~2~ pressure of 16.7 mmHg. A chest X-ray showed mild pulmonary congestion and right pleural effusion. Electrocardiography showed atrial fibrillation and an accelerated idioventricular rhythm, suggesting complete atrioventricular block (Figure [1](#F1){ref-type="fig"}). Transthoracic echocardiography showed moderately impaired left ventricular function with diffuse hypokinesis, but no pericardial effusion or enlargement of the right heart. The end-diastolic left ventricular dimension was 56 mm, the end-systolic left ventricular dimension was 44 mm, and LVEF calculated with the Teichholz formula was 42.9%. The interventricular septal thickness was 11 mm and the posterior left ventricular wall thickness was 12 mm. The diameter of the inferior vena cava was increased to 22 mm, with no respiratory variation. Color Doppler echocardiography showed mild tricuspid regurgitation and the estimated pressure gradient between the right atrium and right ventricle was 25.1 mmHg. Laboratory tests showed neutrophil-dominant leukocytosis, liver dysfunction, and renal insufficiency (Table [1](#T1){ref-type="table"}). His serum creatine phosphokinase (CPK) MB isoenzyme, serum TnI, and plasma BNP levels were markedly elevated. Abdominal ultrasonography showed no abnormalities, other than moderately dilated hepatic veins. Coronary angiography was not performed because of the patient's reduced renal function. The patient developed hemodynamically significant bradycardia (28 beats/min) soon after admission, and a temporary transvenous pacing wire was placed in the right ventricle.
{#F1}
######
Laboratory findings on admission
***Hematology*** **Normal range**
------------------------------------------- -------------------- ---------------------------------
White blood cells\* 13700/mm^3^ 3300 -- 9000/mm^3^
Red blood cells 467 × 10^4^/mm^3^ 430 -- 570 × 10^4^/mm^3^ (Male)
Hemoglobin 13.6 g/dl 13.5 -- 17.5 g/dl (Male)
Hematocrit 41.3% 39.7 -- 52.4% (Male)
Platelets 29.3 × 10^4^/mm^3^ 14.0 -- 34.0 × 10^4^/mm^3^
***Differential white blood cell count***
Segmented neutrophils\* 70.0% 34.0 -- 70.0%
Band cells, stab cells\* 15.0% 1.0 -- 7.0%
Eosinophils 1.0% 0 - 8.0%
Monocytes 5.0% 2.0 -- 10.0%
Lymphocytes\* 9.0% 18.0 -- 49.0%
***Biochemistry***
Total protein\* 6.2 g/dl 6.7 -- 8.3 g/dl
Albumin\* 3.4 g/dl 3.8 -- 5.2 g/dl
AST\* 2660 IU/l 10 -- 40 IU/l
ALT\* 2037 IU/l 5 -- 45 IU/l
LDH\* 3307 IU/l 120 -- 240 IU/l
ALP\* 793 IU/l 100 -- 325 IU/l
γ-GTP\* 197 IU/l 10 -- 50 IU/l
Total bilirubin\* 1.7 mg/dl 0.2 -- 1.2 mg/dl
Direct bilirubin\* 1.0 mg/dl 0 -- 0.4 mg/dl
CPK\* 978 IU/l 60 -- 270 IU/l
CPK-MB\* 231 ng/ml \< 5.2 ng/ml
Blood urea nitrogen\* 63.4 mg/dl 8.0 -- 20.0 mg/dl
Creatinine\* 3.01 mg/dl 0.61 - 1.04 mg/dl
Sodium 141 mEq/l 135 -- 145 mEq/l
Potassium 3.8 mEq/l 3.5 -- 5.0 mEq/l
Chloride 106 mEq/l 98 -- 108 mEq/l
Calcium 8.4 mg/dl 8.4 -- 10.4 mg/dl
Phosphorus 3.1 mg/dl 2.5 -- 4.5 mg/dl
Glucose 99 mg/dl 70 -- 109 mg/dl
C-reactive protein\* 5.50 mg/dl \< 0.30 mg/dl
BNP\* 4806 pg/ml \< 18.4 pg/ml
Troponin-I\* 28.4 ng/ml \< 0.04 ng/ml
***Blood coagulation test***
PT-INR\* 2.19 0.85-1.15
***Serology***
HBsAg Negative Negative
HCV-RNA Negative Negative
IgM anti-HAV Ab Negative Negative
Anti-HIV Ab Negative Negative
\*These values are outside the normal ranges.
*Abbreviations:AST* aspartate aminotransferase, *ALT* alanine aminotransferase, *LDH* lactate dehydrogenase, *ALP* alkaline phosphatase, *γ-GTP* gamma-glutamyltranspeptidase, *CPK* creatine phosphokinase, *BNP* brain natriuretic peptide, *PT-INR* international normalized ratio of prothrombin time, *APTT* activated partial thromboplastin time.
The intravenous administration of dopamine (10 μg/kg/min), unfractionated heparin (15,000 units/day), and furosemide (20 mg/day) was begun on the day of admission. The patient's systolic blood pressure increased to about 120 mmHg and stabilized. The low-dose intravenous administration of human atrial natriuretic peptide (0.025 μg/kg/min) was begun on day 2. Nonsustained monomorphic ventricular tachycardia lasting for about 1 min was observed several times on day 2, but had no significant hemodynamic impact, and was effectively treated with an intravenous bolus injection of amiodarone (150 mg). On day 3, the patient's atrioventricular conduction returned to normal and his daily urine output increased to 3500 mL/day. On day 4, echocardiography showed an improvement in LVEF to 55.2% and the temporary pacing wire was removed. His serum CPK levels peaked on day 4 at 3229 IU/L. On day 7, a chest X-ray showed resolution of the right pleural effusion and his serum creatinine level returned to normal (0.98 mg/dL). The intravenous administration of human atrial natriuretic peptide was stopped, and the patient commenced a regimen of oral losartan (25 mg/day), spironolactone (12.5 mg/day), azosemide (30 mg/day), and warfarin (3.0 mg/day). Supplemental oxygen was no longer needed.
The patient's liver dysfunction improved after admission. His aspartate aminotransferase (AST) level peaked on day 2 (7752 IU/L) and his alanine aminotransferase (ALT) level peaked on day 3 (5899 IU/L) (Figure [2](#F2){ref-type="fig"}). His biliary enzyme levels and total bilirubin also increased, but not as much as his AST and ALT levels. His prolonged prothrombin time returned to normal on day 7. However, he complained of epigastric pain on day 5, and laboratory tests showed increased serum and urine amylase levels. Abdominal computed tomography showed swelling of the pancreatic body and tail, with increased retroperitoneal adipose-tissue density, suggesting pancreatitis (Figure [3](#F3){ref-type="fig"}). No cholelithiasis or tumor occluding the common bile duct or pancreatic duct was observed, and his plasma immunoglobulin G4 level was normal. We administered intravenous gabexate mesilate (600 mg/day) and ceftriaxone (2 g/day) from day 5 to day 12. The patient's abdominal symptoms resolved after a few days, and he recovered without complications. A small amount of melena was observed on day 8, and a total colonoscopy was scheduled. Both his liver dysfunction and pancreatitis had resolved by day 14.
{#F2}
{#F3}
Cardiac catheterization on day 16 showed normal pressure data and a normal cardiac index (3.2 L/min/m^2^). Coronary angiography and left ventriculography showed no abnormalities and normal LVEF (64.2%). A pathological examination of an EMB specimen of the left ventricle showed necrotic cardiomyocytes adjacent to areas with interstitial infiltration of CD3-positive lymphocytes (Figure [4](#F4){ref-type="fig"}A--D), confirming acute myocarditis. The antibody titer against CVA4 increased from 1:4 on day 1 to 1:256 on day 15. On day 19, his plasma BNP level was 238 pg/mL and his serum TnI level was 0.08 ng/mL, and both these values continued to be slightly elevated thereafter.
{#F4}
A colonoscopy on day 18 showed a protruding rectal lesion above the level of the peritoneal reflection. The plasma carcinoembryonic antigen level was 4.7 ng/mL (normal: 0--4.9 ng/mL) and the cancer antigen 19--9 level was 434.6 U/mL (normal: 0--36.9 U/mL). A pathological examination of a biopsy specimen of the rectal lesion showed a moderately differentiated tubular adenocarcinoma invading the muscularis propria. Transesophageal echocardiography on day 27 showed no signs of left atrial thrombus, and the administration of warfarin was stopped in preparation for surgery. The patient was transferred to the Department of Surgery on day 28, and underwent a low anterior resection on day 35. The postoperative diagnosis was stage II rectal cancer (T3N0M0, Union for International Cancer Control Classification). His postoperative recovery was uneventful, with no complications such as heart failure or arrhythmia. He was discharged on day 45. A histological examination of the surgical specimen showed lymphovascular invasion, and adjuvant chemotherapy with folinic acid, fluorouracil, and oxaliplatin was commenced a month after discharge. Interestingly, his plasma BNP level decreased to less than half the preoperative value (78 pg/mL) and his serum TnI level returned to normal (\< 0.04 ng/mL) three weeks after surgery. His plasma BNP level returned to normal (18 pg/mL) four months after surgery. Oral medications for heart failure were stopped five months after discharge. The patient underwent a partial hepatectomy six months after discharge for a solitary metastasis in the liver, but his myocarditis did not recur. Adjuvant chemotherapy was continued until 12 months after surgery. The patient has now been followed-up for two years, with stable LVEF and no recurrence of myocarditis or rectal cancer.
Discussion
==========
Coxsackieviruses belong to the genus *Enterovirus*, in the family *Picornaviridae*, and are important viral pathogens, causing myocarditis and pancreatitis. CVs are subdivided according to their pathogenicity in suckling mice aged less than 48 h into CVA, which causes flaccid paralysis, and CVB, which causes spastic paralysis \[[@B10]-[@B12]\]. There are 23 CVA serotypes and six CVB serotypes \[[@B12]\]. Although CVs are known to cause myocarditis, most cases are attributed to CVB serotypes \[[@B6]\]. CVA infection can cause viral myocarditis, but it has a more benign course than CVB myocarditis \[[@B12]\]. Serotypes A4 and A16 have been most frequently implicated in the rare cases of CVA-associated myocarditis reported \[[@B13]\]. Combined myocarditis and pancreatitis arising from CV infection is very rare, with only three previously reported cases, which were all attributed to CVB infection \[[@B14],[@B15]\]. Ischemic pancreatitis was unlikely in our patient because of the late disease onset (day 5) and hemodynamic stability after admission. Moreover, although we did not stop any medications, including diuretics, the patient's pancreatitis improved spontaneously. Therefore, it is more likely that his pancreatitis developed from his CV infection. To our knowledge, this is the first reported case of an adult patient with myocarditis and pancreatitis attributable to CVA infection. It is unclear why combined myocarditis and pancreatitis is uncommon in CV-infected humans. In a mouse model, the viral titer was higher in the pancreas than in the heart, liver, or spleen during the early stage of CVB infection \[[@B16]\]. In humans, CVB3 capsid protein VP1 was detected in both the cardiomyocytes and islets of endocrine cells in patients who died of acute CVB myocarditis \[[@B17]\]. These results suggest that the rate of latent or asymptomatic pancreatitis may be higher than previously thought in patients with myocarditis attributable to CV infection.
The course of our patient's myocarditis is interesting. Although his LVEF on echocardiography improved soon after admission and normalized on left ventriculography by day 16, an EMB specimen on the same day still showed prominent T-lymphocyte infiltration with a slightly elevated serum TnI level. His serum TnI level had returned to normal three weeks after surgery, and his plasma BNP level had also decreased at this time. This suggests that resection of the rectal cancer may have resulted in the improvement of his myocarditis and inhibited any progression to chronic myocarditis or DCM. The patient's normal LVEF during the two-year follow-up supports this conjecture. The relationship between persistent viral infection and the establishment of chronic myocarditis is controversial. A recent study reported that regulatory T cells (Tregs), which are positive for CD4, CD25, and Foxp3, reduced viral titers and suppress cardiac damage in CVB-infected mice \[[@B18]\]. Furthermore, the myocardial Treg frequency was reduced during the course of CVB3 myocarditis in BALB/c mice and the adoptive transfer of Tregs resulted in the significant amelioration of cardiac fibrosis \[[@B19]\]. Tregs maintain immunohomeostasis and limit autoimmunity, but there may be high levels of Tregs in the blood and tumor tissues of cancer patients, leading to the suppression of antitumor immunity \[[@B20]\]. The proportion of Tregs is significantly higher in colon cancer patients than in healthy controls, and decreases significantly after the radical resection of the tumors \[[@B21]\]. Increased Treg activity also suppresses CD8^+^ T-cell responses to specific antigens of various viruses \[[@B22]\]. The limitations of our report include our inability to determine the myocardial frequency or activity of Tregs before and after surgery, but our findings suggest that the modulation of the immune response associated with rectal cancer and surgery, including a change in Treg activity, may have contributed to the amelioration of myocarditis in our patient. More specifically, the increased Treg activity associated with rectal cancer may have had an inhibitory effect on his cardiac fibrosis and CD8^+^ T-cell-associated myocardial damage during the acute or subacute period of myocarditis, leading to the benign clinical course of this patient, although the onset of myocarditis per se could not be prevented. However, a point of uncertainty persists because reduced Treg activity after surgery might have exacerbated his myocarditis. Despite this, the patient's myocarditis improved. This suggests that components other than Tregs were involved in the improvement in his myocarditis, such as mesenchymal stromal cells (MSCs), which have been identified in the bone marrow and have unique immunoregulatory and regeneration properties \[[@B23]\]. MSCs reduced myocardial inflammation by reducing the expression of proinflammatory cytokines and inducing an increase in Tregs and CD4^+^- and CD8^+^-T-cell apoptosis in CVB3-infected mice \[[@B24]\]. However, it should be noted that no Treg or MSC activity has yet been reported in patients with chronic myocarditis or DCM. Although our patient's clinical course after surgery included the resolution of his myocarditis, further research is required to clarify the mechanisms involved.
Conclusion
==========
We have presented a rare case of myocarditis, liver dysfunction, and pancreatitis arising from CVA4 infection in a patient with undiagnosed advanced rectal cancer. His clinical course was good and his elevated serum TnI and plasma BNP levels returned to normal over four months after the radical resection of his rectal cancer. These features suggest that changes in the immune response to myocarditis that were associated with the patient's rectal cancer may have had a favorable effect on his myocarditis, thereby preventing the progression of acute myocarditis to chronic cardiomyopathy.
Consent
=======
Written informed consent was obtained from the patient for publication of this case report and any accompanying images. A copy of this written consent is available for review by the Editor-in-Chief of this journal.
Abbreviations
=============
CV: Coxsackievirus; EMB: Endomyocardial biopsy; LVEF: Left ventricular ejection fraction; BNP: Brain natriuretic peptide; TnI: Troponin I; CPK: Creatine phosphokinase; AST: Aspartate aminotransferase; ALT: Alanine aminotransferase; DCM: Dilated cardiomyopathy; MSC: Mesenchymal stromal cell; Treg: Regulatory T cell.
Competing interests
===================
The authors declare that they have no competing interests.
Authors' contributions
======================
NA drafted the manuscript. NH and TH collected the patient data and monitored the patient throughout the whole follow-up period. KH, YK, and SS edited the manuscript. MK participated in the study design and coordination and helped to draft the manuscript. All authors have read and approved the final manuscript.
Acknowledgements
================
The authors would like to thank Mr. Yousuke Kitadume for his assistance in preparing the pathological tissue specimen and Miss Hisae Kuribara for her secretarial assistance under the financial support provided by Social Insurance Gunma Chuo General Hospital.
| {
"pile_set_name": "PubMed Central"
} |
{.482}
{.483}
{.484}
{.485}
{.486}
{.487}
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec0005}
===============
Hydatid bone disease is caused by the *Echinococcus granulosus worm*. This zoonosis is endemic in the Mediterranean, Middle East, Asia, Africa and many South American countries, and it is a major public health and economic problem in the Patagonian region of Argentina. Osseous involvement accounts for 0.5 %--4 % of cases in humans. The spine (35 %) and the pelvis (21 %) are the most commonly involved skeletal sites, followed by the long bones, especially the femur (16 %) \[[@bib0005]\]. The location of the disease in the humerus is infrequent. Epidemiological research in some regions has shown: 3 cases of hydatidosis of the humerus between 1971 and 2010 in Serbia, 1 case between 2010 and 2017 in Kazagistan, and no cases have been reported in Spain between 1989 and 2017 \[[@bib0005]\]. Out of 329 cases of hydatidosis reported between 2005 and 2016, only one case of hydatidosis in the humerus was reported in Turkey \[[@bib0010]\].
There is no consensus as regards the medical treatment of hydatidosis in the humerus. Local procedures such as partial resection, curettage, and bone grafting or filling with cement PMMC POLIMETILMETACRILATO were indicated. More severe or extended cases, or where complications such as pathological fractures have occurred, require mega prosthesis or massive allograft for reconstruction \[[@bib0015],[@bib0020]\]. Another radical option is the limb ablation, but the rescue surgery often has better results and is more easily tolerated by the patient, than complete limb amputation, both emotionally and socially \[[@bib0025],[@bib0030]\].
The use of allografts, alloprosthetic composites and prosthetic replacements to reconstruct the proximal humerus after tumor resection and associated complications, have been well described \[[@bib0035]\]. However, the affections of the entire humerus and the recurrences present a challenge of treatment for the reconstruction and rescue of the limb \[[@bib0040]\].
No reports of hydatid disease in the entire humerus have been found in the bibliography. This research presents one case of primary hydatid bone disease affecting the entire humerus, which was treated with radical resection and total endoprosthesis of the humerus.
This work has been reported in line with the SCARE criteria \[[@bib0045]\], and it has been performed in a private academic hospital.
Informed consent was obtained from the patient included in the study, and the institutional Ethical Committee approved the retrospective medical chart review [Fig. 1](#fig0005){ref-type="fig"}.Fig. 1An AP x-rays showing humeral shaft fracture. B MRI showing total humerus compromise.Fig. 1
Case. A 24-year-old patient derived from another center with a diagnosis of delayed healing of left humerus diaphyseal fracture (non-dominant side) with four months of evolution. He had been treated with a brace. At the moment of admission, he had pain and limited motion of his shoulder and elbow. Anteroposterior and lateral radiological images of the humerus showed a line of oblique fracture in the distal third of the diaphysis, and heterogeneous osteolytic and multiloculated images along the entire humerus. The MRI showed pathological images that compromised the entire diaphysis of the humerus and soft tissues at the expansive fracture site, with thinning of the cortical bones. It was decided to perform a biopsy, which resulted in hydatid cysts. The wound of the puncture evolved with the secretion of the hydatids prolonging its healing. Surgical treatment was planned in order to save the limb, given the aggressiveness and expansion of the lesions. Treatment. Oncological resection of the humerus and total replacement of the same with a non-conventional prosthesis designed for the patient was indicated. The surgery was performed using an anteromedial approach extended to the entire diaphysis, achieving total resection. The macroscopic anatomy showed involvement of the entire humerus, including the proximal and distal end with fistulas and vesicles with parasites. The prosthesis at its proximal end consisted of a unipolar head with provisions to attach abductors [Fig. 2](#fig0010){ref-type="fig"}.Fig. 2A--C macroscopic humeral image with proximal, distal, and shaft hydatidosis.Fig. 2
Soft tissue reconstruction was achieved by suturing the divided tendons, i.e., tendons of the rotator cuff, pectoralis major, subscapularis, latissimus dorsi, and teres major, and extensors and flexors in the elbow to the prosthesis. The arm was placed in a sling for six weeks, and the first ROM exercises are started at the elbow. Passive ROM exercises of the shoulder started at four weeks, followed by active exercises at three months and eight weeks after the operation. The patient was also treated with albendazole (15 mg/kg/ d orally) before the operation for one month and after the operation for six months as adjuvant therapy. Result: After two years of follow-up, the patient was pain-free with a range of elbow mobility of 15--90° of flexion-extension, with 0° of pronosupination, and wrist and hand with a full range of motion. At his shoulder he had 30° of anterior flexion, rot ext of 10° and internal rot at the gluteal level.
No local or general disease recurrences were observed. The patient could perform only administrative tasks. The ASES score was 58.33 and Mayo elbow was 65, Quick Dash was 42.3. At present, the patient does not perform tasks or effort with the affected upper limb but can use the hand and the elbow [Fig. 3](#fig0015){ref-type="fig"}, [Fig. 4](#fig0020){ref-type="fig"}.Fig. 3A, B total humerus prosthesis. C postoperative X-rays.Fig. 3Fig. 4Histopatology showing hydatid cysts.Fig. 4
2. Discussion {#sec0010}
=============
Hydatidosis is an endemic disease in different areas of the planet. The primary form in the bones is rare; the humerus has hardly been reported. There is scant information on severe cases with involvement of the entire humerus, including both epiphyses.
The information published following our literature review is case reports with local treatment (curettage), most of them without prolonged follow-up. Surgery is usually the first treatment option. There have been reports of local resections and osteosynthesis, but we believe that in this particular case is not feasible.
Curettage in the humerus has been proposed, but multiple local recurrences have been reported \[[@bib0050]\], as well as the need for multiple reoperations due to local recurrences and loosening of osteosynthesis in fractures. In addition, short follow-ups or lack thereof are reported \[[@bib0055], [@bib0060], [@bib0065]\].
The use of drugs, such as albendazole also seems to be controversial. In an epidemiological study of 17 patients diagnosed with hydatidosis, eight were found with bone involvement, two patients in the spine, 3 in the pelvis, and only 1 in the humerus and 1 in the tibia. None received albendazole, and no treatment or evolution was reported \[[@bib0070]\].
The only real curative procedure seems to be a radical excision. However, it was only possible in 16 percent of the cases \[[@bib0070]\]: when the axial bone-spine, the pelvic bone or the femur are affected.
In many cases, combined therapies based on surgery and antiparasitic drugs are used. However, it is known that the cure is difficult, and recurrence is frequent. We have not identified information on humerus hydatidosis that involves all the bone and its treatment with total resection of the humerus with an endoprosthesis.
The radical resection as salvage of the entire humerus is a challenge. The options are the allograft and the halo prostheses. For both procedures, functional expectations are scarce mainly at shoulder level, functioning as spacers, and to preserve the function of the wrist and hand. In addition to other general complications such as infections and loosening.
Regarding the total endoprostheses, they present functional complications in the shoulder and elbow. There is little experience with tumors, and they are considered a reasonable option in terms of elbow and hand preservation \[[@bib0075]\]. Another problem is usually the migration and dislocations of the prosthesis \[[@bib0080]\]. Natarajan et al· reported the use of a customized megaprosthesis for the reconstruction of bone defects secondary to hydatid bone resection in 3 patients, with excellent results and without recurrences. However, this study had a brief follow-up, which underestimates the likelihood of recurrent infection or loosening of the implant, which complicates prosthetic reconstruction \[[@bib0040]\].
Many authors have reported on the limited active ROM of the shoulder after the proximal humeral and THERs \[[@bib0085],[@bib0090]\]. Careful soft tissue reconstruction with reattachment of the rotator cuff tendons did not seem to improve the results. Despite the poor movement at the shoulder, patients had excellent manual dexterity and functional use of their hands.
The use of tubes and vascular graft for the capsular reconstruction and for the reattachment of the soft tissue in addition to tendon transfers has not shown better results in terms of joint mobility \[[@bib0085],[@bib0095]\].
In the present case, the functional result was not good at the level of the shoulder but functional in the wrist, hand, and elbow, and without recurrence of the disease until the last follow-up. It was planned with that objective.
Recently, techniques with better shoulder results have been reported \[[@bib0085]\]. These authors used a fully expandable humerus prosthesis in a girl with osteosarcoma.
We did not find reports about massive allograft reconstruction in the humerus.
Until the last follow-up of our patient, we did not find glenohumeral subluxation. Marulanda et al. observed no shoulder dislocations in their series of 16 patients with proximal humeral replacement reconstructed with an aortograft mesh. They found that the sleeve created by the aortograft allows for mechanical restraint and facilitates soft tissue reconstruction after tumor resection \[[@bib0100]\]. Puri and Gulia reported a similar experience with no observation of shoulder dislocations in their series of 20 patients with total humeral \[[@bib0105]\].
3. Conclusions {#sec0015}
==============
The cases of hydatidosis in the humerus are infrequent; the management is not standardized, the local resection plus graft and osteosynthesis conserving the bone has shown complications and recurrences. To aim for healing, a radical procedure is necessary with resections and reconstructions with prostheses. Although functional results may be limited, the preservation of the limbs and the function of the hand and wrist are feasible.
Funding {#sec0020}
=======
We did not receive any funding.
Ethical approval {#sec0025}
================
The institutional Ethical Committee approved the retrospective medical chart review.
Consent {#sec0030}
=======
Informed consent was obtained from the patient included in the study, and the institutional Ethical Committee approved the retrospective medical chart review.
Author's contribution {#sec0035}
=====================
Juan Martin Patiño and Alejandro Jose Ramos Vertiz contributed to study design, data collection, data interpretation, and writing.
Registration of research studies {#sec0040}
================================
This manuscript is not a human study, but a case report.
Guarantor {#sec0045}
=========
Juan Martin Patiño.
Provenance and peer review {#sec0050}
==========================
Not commissioned, externally peer-reviewed.
Declaration of Competing Interest
=================================
We do not have conflicts of interest to declare.
This article did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s0005}
============
All living organisms require exogenous foods/nutrients for producing energy in the form of ATP, which is needed for numerous cellular functions. Most of the cellular energy is efficiently produced in specialized organelles mitochondria by oxidative phosphorylation. In addition to energy production, mitochondria play an important role in other cellular functions such as fatty acid oxidation, anti-oxidant defense, intermediary metabolism including ammonia and glutamate detoxification, synthesis of heme and steroids, cell death process, and autophagy [@bib1]. Recent reports showed that mitochondria also undergo constant morphological changes with fusion and fission, following the exposure to toxic agents and/or under pathological conditions [@bib2], [@bib3]. After decreased glucose supply like during fasting or inefficient oxidative phosphorylation in certain disease states, the mitochondrial fat oxidation pathway becomes important in providing an alternative source of energy (e.g., ketone bodies) [@bib1]. Without these energy supply mechanisms due to suppressed mitochondrial function (i.e., mitochondrial dysfunction), living cells would die or become susceptible to cell death processes (necrosis and apoptosis). In fact, it is well-established that mitochondrial function in different tissues is significantly suppressed in numerous medical disorders such as metabolic syndrome including obesity/diabetes, cardiovascular disorders, ischemia reperfusion injury (I/R)^1^ and cancer as well as many neurological disorders like Alzheimer\'s disease and Parkinson\'s disease [@bib4], [@bib5], [@bib6], [@bib7], [@bib8], [@bib9], [@bib10], [@bib11], [@bib12], although etiological causes of each disease are different.
Hepatic mitochondrial abnormalities are observed in both alcoholic fatty liver disease (AFLD) [@bib4], [@bib5], [@bib6], nonalcoholic fatty liver disease (NAFLD) [@bib7], [@bib8], [@bib9] and acute liver injury. As a result, it is conceivable to observe decreased ATP levels and increased fat accumulation (micro-vesicular and macro-vesicular steatosis) in the liver under these conditions. Continued mitochondrial dysfunction with increased oxidative stress also sensitizes the hepatocytes to subsequent necrosis and/or apoptosis of hepatocytes, which likely lead to activation of resident liver macrophage (i.e., Kupffer cells) and recruitment of infiltrating immune cells into the liver with elevated levels of hepatic inflammation (steatohepatitis) and pro-inflammatory cytokines/chemokines [@bib13], [@bib14], [@bib15]. Consequently, hepatic stellate cells can be activated and transformed into myofibroblast-like cells, producing pro-fibrotic cytokines such as transforming growth factor-beta and platelet derived growth factor, leading to hepatic fibrosis/cirrhosis and cancer. These sequential events can be observed in both AFLD [@bib13], [@bib14], [@bib15] and NAFLD [@bib16], [@bib17]. In addition, acute or sub-chronic exposure to various hepatotoxic compounds, including clinically-used drugs, such as acetaminophen (APAP) [@bib18], [@bib19], [@bib20], a major ingredient of Tylenol^®^, an anti-breast cancer agent tamoxifen [@bib21], an anti-retroviral drug zidovudine (AZT) [@bib22] and antidepressants [@bib23], can cause mitochondrial dysfunction, contributing to liver injury with or without fat accumulation, depending on the injurious agent, as extensively reviewed [@bib24], [@bib25], [@bib26]. These hepatotoxic agents and pathological conditions are known to elevate the levels of reactive oxygen and nitrogen species (ROS/RNS) and nitroxidative stress through the suppression of the mitochondrial electron transport chain (ETC) and induction/activation of NADPH oxidase, cytochrome P450 isozymes including ethanol-inducible P450-2E1 (CYP2E1) and CYP4A, xanthine oxidase, and inducible nitric oxide synthase (iNOS). Although increased nitroxidative stress can oxidatively damage mitochondrial DNA and lipids, the majority of insults can also take place at the protein levels through different forms of post-translational modification (PTM) [@bib26], [@bib27]. Because of the newly-emerging information on redox-related protein modifications, we briefly describe five major forms of PTM and functional consequences of some modified mitochondrial proteins in the experimental models of the AFLD, NAFLD and acute liver injury ([Fig. 1](#f0005){ref-type="fig"}). In addition, we highlight potential roles of CYP2E1 in promoting various PTMs.
Post-translational modifications of mitochondrial proteins {#s0010}
==========================================================
Oxidation of mitochondrial proteins {#s0015}
-----------------------------------
Under normal conditions, transiently elevated ROS is known to be involved in cellular signaling pathways [@bib28], [@bib29] and mitochondrial functions are correctly maintained through proper redox balance. However, chronic and/or binge alcohol, high fat diets, tobacco smoking, or certain hepatotoxic drugs can directly damage the mitochondrial ETC, producing greater amounts of ROS leaked from the ETS [@bib24], [@bib25], [@bib26], [@bib27]. Without proper counter-balance by various cellular anti-oxidants, the persistent imbalance in cellular redox states result in decreased levels of mitochondrial antioxidants including mitochondrial glutathione (mtGSH), which serves as a critical determinant between toxic damage and cellular protection [@bib30]. When the cellular defense system is overwhelmed, greater amounts of ROS and RNS remain elevated, ultimately leading to increased nitroxidative stress.
It is well-established that many amino acids such as cysteine (Cys), methionine (Met), histidine (His), proline (Pro), lysine (Lys), tyrosine (Tyr), phenylalanine (Phe), threonine (Thr) and tryptophan (Trp) in most proteins can also be redox regulated. As recently reviewed [@bib26], [@bib27], Cys residue(s) can be oxidatively modified in many forms \[sulfenic acid, disulfide, sulfinic/sulfonic acids, NO- or peroxynitrite-dependent *S*-nitrosylation, NO-independent ADP-ribosylation, mixed disulfide formation between Cys residues and glutathione (glutathionylation), cysteine (cystathionylation), succinic acid, myristic acid [@bib31] or palmitic acid (Cys-palmitoylation) [@bib32] prior to membrane attachment or cellular trafficking\].
Since the sensitive methods to detect redox-regulated Cys residues in cellular proteins were extensively described in our previous reviews [@bib26], [@bib33], we will only highlight the functional consequences of oxidative modification of Cys residues in some proteins. If one of these amino acids serves as the active site or is located near the active site of certain enzymes, it is highly likely that oxidative modifications of these amino acids can result in their inactivation, as shown by the oxidation and/or *S*-nitrosylation of Cys residues including the active site Cys in the mitochondrial aldehyde dehydrogenase (ALDH2) [@bib34] and 3-keto-acyl CoA-thiolase (thiolase) [@bib34] in binge alcohol-exposed rodents. In fact, mass-spectral analysis revealed that more than 87 mitochondrial and 60 cytosolic proteins were oxidatively-modified in alcohol-exposed rodents [@bib34], [@bib35]. The rates of protein oxidation in CYP2E1-containing E47-HepG2 hepatoma cells [@bib36] exhibited a dose- and time- dependent pattern in response to alcohol as well as the presence of CYP2E1 [@bib37]. Similar numbers of oxidatively-modified mitochondrial proteins were also identified by mass-spectral analysis in mice with I/R injury [@bib38] and rats exposed to 3,4-methylenedioxymethamphetamine (MDMA) [@bib39]. The number of oxidatively-modified mitochondrial proteins we characterized may represent a small fraction of the estimated number of 1100--1400 total mitochondrial proteins [@bib40], [@bib41]. However, we believe that the actual number of oxidized proteins could be significantly higher than the proteins originally reported [@bib34], [@bib38], [@bib39], because of the technical limitations in the identification and purification of oxidized proteins and detection methods, especially for those proteins expressed in low abundance, as previously discussed [@bib26], [@bib33].
In the case of oxidative inactivation of mitochondrial ALDH2 through active site Cys modification [@bib42], immunoblot analysis of the immunoaffinity purified ALDH2 proteins by using the specific antibody against ALDH2 or *S*-NO-Cys was performed [@bib34]. Protein detection using the anti-ALDH2 antibody revealed that similar levels of ALDH2 (54 kDa) were observed in all 3 mitochondrial samples \[i.e., mitochondria from control rats, mitochondria from alcohol-exposed rats, and dithiothreitol (DTT)-treated mitochondria from alcohol-exposed rats\]. However, one specific immunoreactive band with the anti-*S*-NO-Cys antibody was detected only in mitochondria from alcohol-exposed rats. Furthermore, the significantly decreased ALDH2 activity in ethanol-exposed rats was fully restored by DDT treatment [@bib34]. Similar results of *S*-nitrosylation and inactivation of ALDH2 were also observed in rat hepatoma H4IIE-C3 cells exposed to NO donors such as *S*-nitrosoglutathione (GSNO), *S*-nitroso-*N*-acetylpenicillamine, and 3-morpholino-sydnonimine, respectively [@bib43]. These results indicate that ALDH2 and other ALDH isozymes, which contain the highly conserved active site Cys residue, can be reversibly modulated by *S*-nitrosylation under increased nitroxidative stress, as reviewed [@bib42]. Other PTMs of Cys and other amino acid residues in ALDH2 protein are summarized ([Fig. 2](#f0010){ref-type="fig"}). Oxidative modification and suppression of ALDH2 in alcohol-exposed rodents were consistently observed by other groups [@bib44], [@bib45]. In fact, these phenomena could explain the decreased levels of ALDH2 activity in alcoholic individuals, as discussed [@bib34], [@bib42]. Oxidative modification and inactivation of ALDH2 were also observed after exposure to MDMA [@bib39] or I/R injury [@bib38], resulting in the accumulation of lipid peroxidation products, which are metabolized by ALDH2 [@bib46]. Similar modifications and inactivation of active site Cys residues in thiolase and possibly other enzymes in the β-oxidation pathway of fatty acids are likely to lead to fat accumulation observed in alcohol-exposed rodents [@bib34], in MDMA-exposed rats [@bib39] as well as in mouse liver following I/R injury [@bib38]. Furthermore, oxidative modification of the active site Cys residue of peroxiredoxin 1 (Prx1) and inactivation was observed in alcohol-exposed mice [@bib47]. Since Prx1 was determined to be localized on the cytosolic side of the endoplasmic reticulum (ER) and interacts with CYP2E1 [@bib47], it is likely that this oxidative inactivation of Prx1 contributes to increased ROS production and fatty liver following alcohol exposure. Based on the conserved redox-sensitive Cys residues among the multiple isoforms [@bib47], [@bib48], [@bib49], many Prx enzymes could be also oxidatively-modified and inactivated, contributing to elevated levels of oxidative stress. Furthermore, the activities of NAD^+^- or NADP^+^-dependent isocitrate dehydrogenases (ICDH) are known to be oxidatively-modified and inactivated in experimental models exposed to hydrogen peroxide, menadione, nitric oxide, or alcohol [@bib50], [@bib51], [@bib52], [@bib53]. Oxidative inactivation of ICDH enzymes would increase oxidative stress since they are important in the regulation of mitochondrial redox balance and cellular defense through providing NADPH for the regeneration of reduced GSH [@bib50]. Mass-spectral analysis confirmed that Cys305 and Cys387 of NADP^+^-ICDH were *S*-nitrosylated, resulting in its inactivation in alcohol-exposed animals [@bib53]. By the same logic, it is expected that many other enzymes that contain active site Cys residue would likely undergo similar types of redox-related modifications and become inactivated, as reviewed [@bib26], [@bib33]. For instance, it was reported that a DNA repair enzyme O^6^-methylguanine-DNA-methyltransferase which contains redox-sensitive Cys in its active site can be inactivated through *S*-nitrosylation [@bib54], resulting in increased levels of oxidative DNA modifications, which can contribute to carcinogenesis.
It is also possible that redox regulation can indirectly affect the patterns of gene expression by modulating the conserved Cys residues in proteins involved in the cell signaling pathways like many protein phosphatases (see the Section "Phosphorylation of mitochondrial proteins") or directly affecting many redox-sensitive transcription factors such as nuclear factor-kappa B (NF-κB), nuclear factor (erythroid-derived 2)-like 2 (Nrf2) and hypoxia-inducible factor-1 (HIF-1), as recently reviewed [@bib55]. The modulation of transcription factors can either be harmful or beneficial. For instance, NF-κB, a critical transcription factor in inflammation and innate immune regulation, can be activated by alcohol [@bib56] and other toxic agents such as lipopolysaccharide (LPS) [@bib57]. NF-κB can be activated through proteasomal degradation of the inhibitor protein (I-κB) of the NF-κB after its phosphorylation by I-κB kinase. Consequently, NF-κB activation leads to the transcriptional initiation of many pro-inflammatory cytokines, chemokines and pro-oxidant proteins such as TNF-α, IL-12, monocyte chemotactic protein-1, macrophage inflammatory protein-2, cyclooxygenase-2 and iNOS [@bib55], [@bib56], [@bib57]. Chronic or acute alcohol exposure and other hepatotoxic agents can also activate HIF-1α in mice and rats [@bib57], [@bib58], [@bib59], [@bib60], [@bib61]. However, activation of HIF-1α, which is co-localized with CYP2E1 in centrilobular areas, seems to contribute to CYP2E1-dependent oxidative stress and liver toxicity, as recently suggested in chronic and acute alcohol exposure [@bib59], [@bib60], [@bib61]. In contrast, a cell-protective transcription factor Nrf2 can be activated under increased oxidative stress in AFLD and NAFLD [@bib62], [@bib63], [@bib64], [@bib65]. Increased oxidative stress activates Nrf2 through oxidative modifications of redox-sensitive Cys residues of the regulatory binding protein Kelch-like ECH-associated protein 1 (Keap1), leading to ubiquitin-dependent degradation of Keap1, promoting Nrf2 translocation into the nucleus for transcriptional activation of its downstream targets, as reviewed [@bib66]. Activation of Nrf2 with its down-stream target proteins, including glutamate cysteine ligase, heme oxygenase-1, quinone reductase, and thioredoxin reductase, provides anti-oxidant defense against increased oxidative stress caused by alcohol exposure [@bib62].
As described thus far, oxidative modifications of Cys residues in many cellular proteins are associated with inactivation or suppression of the activity of modified target proteins. However, redox modification of Cys, especially the mixed disulfide formation, such as *S*-glutathionylation, can also be considered as a protective mechanism in some cases. For instance, reversible oxidation of Cys residues of mitochondrial NADH-dependent ubiquinone oxidoreductase (complex I), ALDH2, and Prx3 seem to be protective, despite temporary inactivation, against further tissue damage from irreversible hyper-oxidation of Cys to Cys-sulfinic acid (Cys-SO~2~) and Cys-sulfonic acid (Cys-SO~3~) followed by proteolytic degradation [@bib67], [@bib68], as in the case of I/R injury [@bib69], [@bib70], [@bib71], [@bib72]. It is known that Cys-glutathionylation of mixed disulfide groups can create a functional cycle consisting of the forward reaction (glutathione S-transferase P) and the reverse reaction (glutaredoxin system) [@bib67], [@bib73], [@bib74]. Furthermore, some Cys residues can be modified through adduct formation with carbonyl compounds, including lipid peroxide 4-hydroxy-2-nonenal (4-HNE) and reactive metabolites of APAP. However, Cys residues in the adducts with reactive metabolites or lipid peroxides cannot be restored even in the presence of a strong reducing agent DTT and thus are considered irreversible, as discussed [@bib26], [@bib33]. The functional implications of some examples of these adducts are described later (see "Section Electrophilic adduct formation of mitochondrial proteins").
Nitration of many mitochondrial proteins {#s0020}
----------------------------------------
It is well-established that the physiological levels of NO produced by endothelial NOS (eNOS) and neuronal NOS (nNOS) are relatively small and can act as a signaling molecule within the cell or interact with surrounding cells [@bib27], [@bib75], [@bib76]. However, many toxic agents such as alcohol, APAP and LPS can induce iNOS, which produces greater amounts of NO, likely leading to nitrosative/nitrative stress [@bib27], [@bib75], [@bib76], [@bib77]. In this case, NO can interact with ROS to produce a potently toxic peroxynitrite, which can nitrate Tyr residues to generate 3-nitroTyr (3-NT) or *S*-nitrosylate Cys residues of many proteins [@bib77]. These two PTMs can lead to alterations of both protein structure and function, contributing to increased cell and tissue injury.
Protein nitration may occur in various cell compartments. Although the concentration of mtGSH is similar to that of cytosolic GSH (5--10 mM), it represents a small fraction of the total cellular GSH [@bib30]. Because mitochondria are a major source of ROS/RNS as previously mentioned, they can be very vulnerable to nitroxidative damage [@bib27]. Indeed, mitochondrial DNA is extremely sensitive to oxidative/nitrative damage due to (A) its location closer to the inner mitochondrial membrane; (B) the relatively low amount of histone, anti-oxidant enzymes, and mtGSH, and (C) the low levels of polyamines and DNA repair enzymes [@bib78], [@bib79]. In support of the view of mitochondrial sensitivity to nitration, it has been reported that the turn-over rates of rat liver mitochondria proteins were markedly decreased from few days to hours as a result of proteolytic degradation following exposure to nitrating conditions [@bib80], [@bib81]. Interestingly, mitochondrial DNA encodes 13 polypeptides all of which are members of the 4 mitochondrial ETC proteins, namely Complexes I, III, IV, and V [@bib82], [@bib83]. Taken together, mitochondrial DNA mutation and/or deletion may lead to abnormal expression or deletion of some of the ETC components, contributing to mitochondrial respiratory dysfunction and more leakage of ROS/RNS. Consequently, mitochondrial damage or dysfunction due to oxidative/nitrative protein modifications can be expected in various pathological conditions or following exposure to some toxic agents such as alcohol, high fat, MDMA, and APAP [@bib18], [@bib19], [@bib20], [@bib21], [@bib22], [@bib23], [@bib24], [@bib84], [@bib85]. Nitration of mitochondrial proteins and their functional roles in fatty liver disease and/or acute liver injury have been extensively reviewed recently [@bib27]. For instance, NO-dependent inhibition of the mitochondrial respiration was monitored in both AFLD and NAFLD [@bib86], [@bib87], [@bib88], [@bib89], [@bib90].
Since protein nitration is one of the emerging research topics in acute tissue injury, we specifically focused on this PTM in promoting mitochondrial dysfunction and acute hepatotoxicty. We used APAP as a well-established model agent for acute drug-induced liver injury (DILI) to study the functional role of nitrated mitochondrial proteins and the mechanisms of mitochondrial dysfunction and hepatotoxicity. APAP was first shown to induce and promote mitochondrial protein nitration and that peroxynitrite can cause mitochondrial DNA damage by Cover et al. [@bib18]. Hepatic necrosis highly correlated with the areas of nitrated proteins [@bib91]. Direct evidence for a critical role of peroxynitrite in APAP-mediating hepatic injury has been provided with the seminal work by Knight et al. [@bib92]. This study showed that the protective effects of GSH were mainly due to the restoration of cellular GSH levels for efficient scavenging of peroxynitrite, thus providing direct evidence of the central role of peroxynitrite in APAP-induced liver injury. CYP2E1, one of the main enzymes involved in APAP metabolism, seems to play a critical role in APAP-induced protein nitration and liver injury [@bib93]. Based on these results, we hypothesized that CYP2E1-dependent nitration of many mitochondrial proteins is important in promoting mitochondrial dysfunction and liver injury. In fact, after immunoaffinity purification followed by mass-spectral analysis, we demonstrated that more than 70 mitochondrial proteins and 30 cytosolic proteins were nitrated in APAP-exposed mice [@bib94]. These results are in line with another recent report [@bib95]. The identified nitrated proteins are involved in cellular energy production, fat metabolism, intermediary metabolism, urea cycle, chaperone activity, and anti-oxidant defense. To further determine the functional roles of protein nitration, we mainly focused on the five selected mitochondrial enzymes, namely SOD2, Gpx, ATP synthase, ALDH2 and thiolase. We specifically studied the presence of nitration and activity change in each of these five enzymes at 2 h after exposure to a toxic dose of APAP in the presence or absence of the co-treatment with an anti-oxidant *N*-acetyl-cysteine (NAC) [@bib94]. Treatment with NAC completely protected the liver from the APAP-induced necrotic injury measured by increased serum transaminase activities and liver histology. Furthermore, NAC ameliorated the inactivation of the five critical enzymes via preventing Tyr nitration, as evidenced by immunoprecipitation with a specific antibody to each protein followed by immunoblot with anti-3-NT antibody. Thus, protein nitration serves as one of the critical causal factors in APAP-induced acute liver damage usually observed at later time points. Indeed, the suppressed activities of anti-oxidant enzymes such as SOD2 and Gpx would certainly interfere with the cellular anti-oxidant abilities. Suppressed ATP synthase activity would lead to decreased ATP production, interfering with vital cell functions and ultimately causing necrosis. Furthermore, inactivation of thiolase likely leads to the inhibition of the β-oxidation of fatty acids with increased hepatic fat accumulation or inefficient supply of an alternative energy (e.g., ketone bodies produced from fat degradation). Furthermore, suppressed ALDH2 would lead to the accumulation of toxic, reactive aldehydes and lipid peroxidation, as previously mentioned in the oxidation section [@bib42]. Based on our results, we expect similar results of mitochondrial dysfunction and tissue injury with other conditions such as I/R injury [@bib38] and LPS treatment [@bib96], which caused extensive protein nitration. It is important to mention that the inactivation of many nitrated proteins can result from the nitration of Tyr residue in the active site, leading to interference of the protein's catalytic activity. However, it is still possible that nitration of other Tyr residues may cause a conformational change in the target protein and thus block the accessibility of ATP or substrates for example, as previously discussed [@bib27]. However, it would be prudent to double-check the function of each nitrated protein, since nitration can also increase the activity of some proteins such as glutathions S-transferase (GST), heat shock protein 90 (HSP90), and protein phosphatase type 2 \[PP2\], or exert no effects at all as reported with transferrin [@bib27].
Although most reported results mainly focused on Tyr nitration, nitration of Trp residues in proteins was also reported [@bib97]. However, there have been fewer reports about Trp nitration and functional consequence than those of Tyr nitration, since Trp is much less abundant than Tyr and that Trp is usually buried inside the protein core with less surface exposure for nitration. Nonetheless, excellent reports discussed the occurrence of Trp nitration that may cause functional alterations in detail [@bib97], [@bib98], [@bib99]. The effects of peroxynitrite on nitration of Tyr or Trp and *S*-nitrosylation of Cys likely depend on cellular status of antioxidants, various microenvironments including surface exposure, peptide loop structure, local charge groups, and the presence of competing amino acids nearby (e.g., Cys near the potentially-nitrated Trp or Tyr), as suggested [@bib98].
Phosphorylation of mitochondrial proteins {#s0025}
-----------------------------------------
Many cellular proteins can be regulated by reversible cycles of phosphorylation and dephosphorylation by protein kinases and phosphatases, respectively. It is well-established that many exogenous agents such as alcohol, smoking, high fat, and APAP can modulate many different classes of protein kinases including cAMP-dependent protein kinases (PKA), PI3K/AKT/mTOR (PKB), protein kinase C isoforms (PKC), calcium-calmodulin-dependent protein kinases, stress activated mitogen-activated protein kinases (MAPKs) including c-Jun N-terminal protein kinase (JNK) and p38 kinase (p38K), and receptor-mediated tyrosine kinases. These kinases can work together synergistically or antagonize each other's activities. In fact, crosstalk between different classes of protein kinases becomes complicated, contributing to different outcomes, depending on their inducibility, tissue distribution, subcellular localization or translocation to another compartment, cellular environment, treatment agent, and activity status of phosphoprotein phosphatase, as recently reviewed [@bib100], [@bib101], [@bib102], [@bib103], [@bib104], [@bib105]. For instance, concurrent activation of PKC-delta isoform along with stress activated protein kinases generally enhances cell death [@bib106], [@bib107], [@bib108], although some exceptions exist. In contrast, activation of PI3K/Akt and extracellular signal regulated protein kinase (ERK) is generally known to support cell survival and proliferation. However, we briefly describe the redox-mediated regulation of stress activated MAPKs and AMP-activated protein kinase (AMPK) in this review, based on their involvement in mitochondrial homeostasis, intermediary metabolism for steatosis and liver injury.
Under increased nitroxidative stress, stress-activated MAPKs can be stimulated partly through inhibition of phosphoprotein phosphatases. For instance, the highly-conserved active site Cys residue of phosphoprotein phosphatases including MAPK phosphatases [@bib103], [@bib104], [@bib105] can be oxidatively-modified and inactivated by the mechanisms described earlier ("Section of Oxidation of mitochondrial proteins"). This redox-related activation of MAPKs is similar to that of Akt activation through oxidative inactivation of the essential Cys residue of its binding partner lipid phosphatase and tensin homolog (PTEN) [@bib109]. The inactivation of MAPK-phosphatases leads to activation of various stress-activated JNK and p38K [@bib104], [@bib105]. In fact, acute exposure to many hepatotoxic agents including alcohol [@bib110], [@bib111], APAP [@bib112], [@bib113], [@bib114], [@bib115], [@bib116], arachidonic acid [@bib117], 4-HNE and carbon tetrachloride [@bib118], palmitic acid [@bib119], and troglitazone [@bib120] significantly activated JNK (p-JNK) and/or p38K (p-p38K), contributing to toxicity in a variety of cells. Pretreatment with a specific inhibitor of CYP2E1 blocked metabolic activation of some hepatotoxic agents, which are CYP2E1 substrates [@bib110], [@bib111], [@bib112], [@bib113], [@bib114], [@bib115], [@bib116], [@bib117], [@bib118], and prevented JNK activation. Taken together, CYP2E1 is critically involved in JNK activation and cell death caused by those hepatotoxic CYP2E1 substrates. In addition, activation of JNK was positively related to high fat diet-induced insulin resistance and metabolic syndrome [@bib119], [@bib121], [@bib122], [@bib123]. A recent study also showed that sustained JNK activation led to phosphorylation at Ser-46 of sirtuin 1, NAD^+^-dependent deacetylase in cytosol (described later), followed by ubiquitin-dependent degradation, contributing to fat accumulation in obese mice [@bib124]. Furthermore, persistent activation of JNK and/or p38K is positively associated with various liver diseases such as steatohepatitis and I/R injury [@bib122], [@bib123], [@bib124], [@bib125], suggesting the clinical importance of JNK and p38K in human disease states, as reviewed [@bib126]. Experiments with gene deletion technology and anti-sense oligonucleotides were conducted to determine the role of JNK1 versus JNK2 in hepatotoxicity. For instance, both JNK1 and JNK2 seem to be important in alcohol-induced fatty liver injury [@bib111], while JNK1 seems critical in hepatotoxicity caused by co-treatment of pyrazole and TNFα [@bib127]. In contrast, JNK2 plays a dominant role in APAP-induced acute liver injury since *JNK2-null* mice are resistant to acute hepatotoxicity [@bib113], suggesting that this area needs further clarification.
Several reports on the redox-related activation of the JNK and/or p38K or their upstream kinases such as mitogen activated protein kinase kinase (M2K) and mitogen activated protein kinase kinase kinase (M3K), as reviewed [@bib128]. However, the number and identity of the mitochondrial and cytosolic proteins, that are phosphorylated by active p-JNK and/or p-p38K, need to be investigated. Our results showed that pro-apoptotic Bax in the cytosol can be phosphorylated (activated) by p-JNK and/or p-p38K, resulting in its conformational change with exposure of the C-terminal membrane domain. These events were followed by translocation of the activated Bax to mitochondria, leading to changes in mitochondrial permeability transition (MPT) and apoptosis of cultured hepatoma cells [@bib129]. The importance of phosphorylation was further confirmed using site-directed mutagenesis of potential amino acids for phosphorylation followed by functional analysis. Our results revealed that Thr167 of Bax was phosphorylated before it was translocated to mitochondria to stimulate mitochondria-dependent apoptosis [@bib129]. In addition, activated p-JNK is known to translocate to mitochondria to initiate MPT change and damage following exposure to hepatotoxic agents including APAP [@bib115], [@bib116], [@bib130], [@bib131]. Since JNK does not contain the canonical mitochondrial leader sequence [@bib132], it would be interesting to identify the mechanism(s) by which activated p-JNK is transported into mitochondria and then phosphorylates mitochondrial matrix proteins such as ALDH2 in CCl4-exposed rats [@bib133]. One of the potential mechanisms of JNK translocation to mitochondria could be through its interaction with a scaffold protein Sab [@bib116], [@bib134] or Bcl-XL [@bib130] on the mitochondrial outer membrane before it gets transported into mitochondria. When the active p-JNK is translocated to mitochondria, it can phosphorylate many mitochondrial proteins such as ATP synthase β subunit, α-KGDH, PDH E1α and β subunits, which were phosphorylated by the recombinant JNK, as shown in an in vitro study using bioenergetically competent mitochondria [@bib135]. However, the identities and physiological roles of numerous mitochondrial proteins phosphorylated by p-JNK (e.g., JNK-specific phosphoproteomes) need to be determined in animal models and human disease states in the future.
Regulatory mechanism of AMP-dependent protein kinase (AMPK) by alcohol and other non-alcoholic substances, including high fat diet has become an important research topic, since AMPK serves as a key sensor and/regulator of metabolic syndrome and many cellular processes [@bib136]. In fact, the redox sensitive AMPK can be paradoxically regulated by activation or suppression. For instance, AMPK can be activated by hydrogen peroxide and mitochondria-derived ROS since its activation can be prevented by pretreatment with antioxidants such as NAC or GSH-ethyl ester [@bib101]. AMPK activation by hydrogen peroxide was tightly associated with the elevated levels of AMP with concurrent depletion of ATP. Recent results showed that AMPK can be also activated in an AMP-independent manner. Regardless of AMP dependency, AMPK activation under oxidative stress or by chemical activators such as metformin and AICAR can lead to increased cell survival signal by promoting autophagy to remove damaged mitochondria and to stimulate mitochondrial energetics [@bib101]. In contrast, AMPK can be suppressed by alcohol [@bib137] or other agents such as high fat diet [@bib138] and APAP [@bib139]. Inactivation of AMPK is known to be associated with AFLD and NAFLD or acute liver injury, as reviewed [@bib13].
Acetylation of mitochondrial proteins {#s0030}
-------------------------------------
Reversible protein acetylation and deacetylation, catalyzed by acetyltransferases and deacetylases, respectively, represent another form of PTMs, although functional roles of many acetylated proteins are still unclear and need further investigations. In fact, acetylation on *N*-ε-Lys residues in proteins becomes an important research area through epigenetic regulation of many genes involved in normal cellular growth and physiology by acetylated histones and other chromatin associated proteins [@bib140]. A large scale proteomics study revealed that approximately 20% of mitochondrial proteins and 44% of NAD^+^-dependent mitochondrial dehydrogenases are acetylated under physiological conditions [@bib141]. The number of acetylated proteins can be significantly increased in experimental models of pathological conditions or following exposure to hepatotoxic agents, possibly resulting from redox-mediated suppression of deacetylases, as shown in alcohol-exposed [@bib142], [@bib143], [@bib144], [@bib145] or high-fat exposed rodents [@bib146], [@bib147], [@bib148]. For instance, chronic and binge alcohol exposure suppressed the activities or decreased levels of various isoforms of sirtuin, Zn and NAD^+^-dependent deacetylases, involved in cellular aging, lipid metabolism, and anti-oxidant defense. Because of NAD^+^-dependence, the activities of sirtuins can be suppressed by the redox change (e.g., a decreased NAD^+^/NADH ratio) following alcohol metabolism via alcohol dehydrogenase (ADH) and low *K*~m~ ALDH2, both of which use NAD^+^ as a cofactor for their activities. Earlier reports also showed that chronic alcohol exposure suppressed the activities and protein amounts of sirtuin 1, 3 and 5 isoforms [@bib142], [@bib143], [@bib144], [@bib145], [@bib149], [@bib150], in a CYP2E1-independent manner [@bib144]. In addition, the activities of sirtuin isoforms can be allosterically suppressed through thiol-specific modification of Cys280 in the Zn binding site with 4-HNE adduct formation, as demonstrated [@bib151]. Consequently, the levels of acetylated proteins were significantly elevated in alcohol-exposed rodents. Some of the acetylated proteins, that are deacetylated by sirtuin isoforms, are several transcription factors involved in the regulation of lipid metabolism and cell apoptosis such as sterol regulated element binding protein (SREBP), peroxisomal proliferator activated receptor gamma coactivator-1 (PGC-1), forkhead box protein O1 (FoxO1), NF-κB and p53 [@bib142], [@bib143], [@bib144], [@bib149]. Acetylation of these transcription factors is a well-established cause for fat accumulation and altered cell proliferation in the liver. For instance, alcohol-related suppression of sirtuin 1 increased the levels of acetylated SREBP-1 and PGC-1α with functional activation and suppression, respectively, resulting in alcoholic fatty liver [@bib142]. By using immunoprecipitation with anti-Ac-Lys antibody followed by mass spectral analysis, 91 acetylated mitochondrial proteins were recently identified in alcohol exposed mice [@bib152]. Some of the acetylated mitochondrial proteins are ALDH2, ATP synthase, Gpx, and thiolase, suggesting that these acetylated mitochondrial proteins likely contribute to their functional alterations and fat accumulation [@bib152]. Another report further indicated that alcohol treatment elevated the number of acetylated proteins as well as propionylated protein in mitochondria and nuclear fractions with little changes in cytosolic and microsomal fractions through suppression of sirtuin activities [@bib153]. However, in this study, alcohol-exposure did not seem to change the protein levels of sirtuin 3, 4, and 5. The reasons for dissimilar results on sirtuin levels, observed in different laboratories, are unknown, but could be due to distinct ethanol dose, time course, environment, or animal strains, and warrants further investigation.
In addition, high fat feeding can suppress the activities of mitochondrial sirtuin 3 and other isoforms, thus increasing the number of acetylated proteins. However, the mechanism(s) by which high fat decreased levels of sirtuin isoforms is poorly understood. The levels of hyper-acetylated proteins are markedly increased in mice deficient of mitochondrial sirtuin 3 [@bib147], [@bib154], [@bib155]. For instance, acetylation of long-chain fatty acyl-CoA dehydrogenase, one of the four enzymes involved in the mitochondrial fat oxidation pathway, causes its inactivation, leading to fat accumulation [@bib154], [@bib155]. In contrast, activation of sirtuins by small molecule polyphenols including resveratrol [@bib156] and green tea extracts [@bib157] can reduce the number of acetylated proteins, including PGC-1α, which contributes to increased fat oxidation and insulin sensitivity with improved outcome against metabolic syndrome. Therefore, sirtuin 3 and its isoforms, involved in fat metabolism and aging process, become important therapeutic targets in managing metabolic syndrome and cellular senescence in the liver and other tissues.
As mentioned earlier, key mitochondrial proteins in regulating the fatty acid metabolism are acetylated and that the number of acetylated proteins can be significantly increased through suppression of the mitochondrial deacetylase sirtuin 3 and other sirtuin isoforms by alcohol [@bib142], [@bib143], [@bib144], [@bib145], [@bib149], [@bib150], [@bib151], [@bib152], [@bib153], high fat [@bib148], and other conditions [@bib154], [@bib155]. Compared to numerous reports on the regulation of deacetylase proteins including mitochondrial sirtuin 3, the expression and properties of the mitochondrial acetyltransferase enzyme(s) have been less characterized although acetylation sites in many mitochondrial proteins have been recently reported [@bib158]. In addition, other studies indicated that protein acetylation can take place nonenzymatically in the presence of low levels of acetyl-CoA [@bib159] or acetyl-phosphate and acetyl-adenylate [@bib160]. These reports indicate that the characteristics of acetyltransferase(s) are poorly understood and need further studies, although the GCN5 family seems to be the prototype of (histone) acetyltransferases [@bib161].
Electrophilic adduct formation of mitochondrial proteins {#s0035}
--------------------------------------------------------
It is well-established that excessive alcohol intake and many potentially toxic compounds can suppress low *K*~m~ mitochondrial ALDH2 activity in rodents and humans through various PTMs of the active site Cys residue and other redox-sensitive amino acids, as reviewed [@bib26], [@bib27], [@bib42], resulting in the accumulation of toxic acetaldehyde and other lipid aldehydes. Consequently, cellular macromolecules including mitochondrial proteins [@bib162], [@bib163] and DNA [@bib164] are subjected to modification by these toxic aldehydes, leading to mitochondrial dysfunction and oral and gastrointestinal tract cancers [@bib165]. Chronic alcohol drinking and/or smoking or high fat can also cause mitochondrial damage with increased nitroxidative stress and lipid peroxidation products such as 4-HNE, 4-oxonon-2-enal \[4-ONE\], malondialdehyde (MDA), and acrolein (ACR) [@bib166], [@bib167], [@bib168]. In addition, many potentially toxic drugs can produce reactive metabolites, which can induce mitochondrial dysfunction and acute liver injury [@bib18], [@bib19], [@bib20], [@bib21], [@bib24]. For instance, APAP metabolism produces *N*-acetyl-*p*-benzoquinone imine (NAPQI) while the metabolism of potentially hepatotoxic troglitazone generates reactive quinolone-like metabolites. Other drugs such as halothane, tienilic acid, and dihydralazine can produce reactive metabolites, which promote hepatotoxicity. The underlying mechanisms of these idiosyncratic toxicities by these agents could be due to interactions between highly reactive electrophilic intermediates with various mitochondrial proteins, leading to mitochondrial dysfunction and tissue damage including DILI. They can also damage membrane integrity and alter calcium homeostasis between endoplasmic reticulum and mitochondria, leading to hepatotoxicity. Furthermore, these protein adducts can serve as a neo-antigen of hapten--carrier conjugate proteins to activate immune responses and to facilitate infiltration of neutrophils into the liver to promote inflammation [@bib18], [@bib19], [@bib20], [@bib21], [@bib24], [@bib166], [@bib167], [@bib168].
Recent results showed that highly-reactive lipid peroxides including 4-HNE and 4-ONE, despite their extremely short half-lives, can suppress the activities of many mitochondrial proteins likely through adduct formation with a few amino acids such as Cys, His, Arg, and Lys of target proteins [@bib168]. For instance, earlier reports showed that acetaldehyde and 4-HNE produced by alcohol intake interact with mitochondrial proteins, including cytochrome C oxidase (complex IV), leading to its inactivation, mitochondrial dysfunction and liver injury [@bib169], [@bib170], [@bib171], [@bib172]. By using mass-spectrometry, Petersen and colleagues recently determined the specific amino acids that are modified by 4-HNE. They showed that the active site Cys301 of ALDH2 [@bib173] and Cys280 in the critical Zn binding site of sirtuin 3 [@bib151] were covalently modified with 4-HNE, leading to their inactivation. In addition, many other proteins such as protein disulfide isomerase [@bib174] and Grp78 [@bib175] in the ER, gamma-glutamyl-cysteine synthase (GCS) [@bib176], ERK [@bib177], PTEN [@bib178], or AMPK [@bib179] in cytosol could also be inactivated through interaction with 4-HNE, likely leading to ER stress, GSH depletion and alteration of the cell signaling pathways frequently observed in alcohol-fed models and humans.
Unlike mixed disulfide bonds of Cys such as *S*-nitrosylation and *S*-glutathionylation, the protein adducts with reactive carbonyls including 4-HNE, 4-ONE, MGO and ACR cannot be reversed by the presence of reducing agents such as DTT, GSH and ascorbate, as recently reviewed [@bib28]. Therefore, these protein adducts may exist for an extended time period unless they are removed by cellular proteolytic enzymes or phagocytic macrophages. Furthermore, HNE-protein adducts can induce immune cell activation of hepatic macrophage Kupffer cells and/or infiltration of neutrophils into the liver, producing inflammatory cytokines and activation of stellate cells, leading to advanced liver disease, as reported in alcoholic liver injury [@bib166], [@bib167], [@bib168], [@bib180]. In fact, the average levels of MDA- and HNE-protein adducts in 50 alcoholic cirrhosis patients were significantly higher than those in a similar number of patients with nonalcoholic cirrhosis or healthy individuals [@bib180]. In contrast, no differences in the levels of ACR- and MGO-protein adducts were found among different groups. Furthermore, the average levels of MDA and HNE-adducts were higher in 51 alcoholic patients with advanced liver disease with fibrosis/cirrhosis than 23 fatty liver disease or 30 healthy control people, suggesting that MDA- and HNE-protein adducts likely served as antigens or activators of immune responses in advanced alcoholic liver disease. The ethanol-induced CYP2E1 was partly responsible for the production of 37 kD-acetaldehyde protein adducts [@bib181] while 4-HNE protein adduct could be produced in CYP2E1-independent manner, although CYP2E1 seems to at least play a permissive role [@bib182].
In case of high-fat fed rodents, the total number of the reports on HNE-protein adducts in the livers of NAFLD seems relatively small compared to those in AFLD, although the identities of HNE-protein adducts were reported in extra-hepatic tissues such as heart [@bib183], [@bib184]. Mass spectral analysis revealed 39 unique lipoxidation sites in 27 mitochondrial proteins. These lipid peroxide-modified mitochondrial proteins include: pyruvate dehydrogenase, malate dehydrogenase, ICDH, methylmalonate-semialdehyde dehydrogenase, long chain fatty acyl-CoA dehydrogenase, NADH-dependent ubiquinone oxidoreductase (complex I), succinate dehydrogenase (complex II), ubiquinone-cytochrome bc1 oxidoreductase (complex III), cytochrome c oxidase (complex IV), ATP synthase (complex V), and others [@bib184]. Although the activities of each HNE-modified protein were not determined in this study, it is reasonable to assume that the functional activities of some of the modified proteins could be suppressed. We also believe similar types of lipid peroxide adducts present in the liver of high-fat exposed rodents (and humans), due to the elevated levels of oxidative stress and lipid peroxides including 4-HNE [@bib185], as comparable to the alcoholic fatty liver [@bib170], [@bib171], [@bib172]. In fact, ALDH2 activity in the heart was significantly inhibited in high fat-exposed diabetic mice [@bib186]. Further characterization revealed that ALDH2 was inactivated through 4-HNE adduct formation, as evidenced by immunodetection of 4-HNE adduct in the immunoprecipitated ALDH2 with the anti-ALDH2 antibody [@bib186]. The levels of cardiomyocyte hypertrophy and dysfunction in high fat-exposed mice were inversely correlated with the ALDH2 activity, suggesting a causal role of ALDH2 in diabetic cardiomyopathy. These results are consistent with the cardioprotective role of ALDH2 which can be activated by small molecule activators [@bib42], [@bib187] or over-expression by genetic modulation [@bib188], [@bib189].
In addition to lipid peroxide-related protein adducts, recent data suggest that sugar moiety-related protein adduct such as advanced glycation end products (AGEs) can be produced by high fat or excessive alcohol exposure or under pathological states such as diabetes, cardiovascular diseases and Alzheimer\'s disease, as recently reviewed [@bib190], [@bib191], [@bib192], [@bib193]. Elevated levels of cross-linked AGEs in many tissues can up-regulate the expression of the receptor for AGEs (RAGE), which activates NADPH oxidase, leading to increased oxidative stress, insulin resistance, and inflammation with disruption of normal cell signaling pathways. These pro-oxidative events with accumulated proteins contribute to acute injury and/or persistent chronic disease states in many tissues [@bib190], [@bib191], [@bib192], [@bib193]. For instance, chronic alcohol exposure elevated the levels of acetaldehyde-derived AGE (AA-AGE) and HNE-protein adducts in the livers of rats and humans while their elevated levels disappeared after alcohol abstinence. The immunostaining intensities of AA-AGE and HNE-adducts positively correlated with the severity of alcoholic liver disease [@bib193]. Furthermore, AA-AGE but not *N*-ethyllysine (NEL) caused apoptosis of rat hepatocytes, suggesting an important role of AA-AGE in alcohol-induced hepatotoxicity and fatty liver injury [@bib193]. Similar results of elevated production of AGE and its plasma levels in cirrhotic patients compared to those in healthy individuals were previously reported [@bib194]. High fat feeding also elevated the levels of AGEs and RAGE in many tissues, including kidneys, hearts and epididymal fat in an aldose reductase-dependent manner [@bib195]. In a mouse model of NASH caused by high fat western diet or choline-deficient diet, AGEs induced TNFα-cleavage enzyme in an NADPH oxidase-dependent manner and stimulated fibrogenic activity via decreasing the levels of sirtuin 1 and tissue inhibitor of metalloproteinase 3 (TIMP3) [@bib196]. The significantly lower levels of sirtuin 1 and TIMP3 were also observed in livers from NASH patients than those of healthy individuals, suggesting a role of AGE adducts in promoting inflammatory responses. Another report recently indicated that several mitochondrial proteins in aged rats were modified with AGEs, thus their activities inactivated. These AGE-modified proteins were: complexes I and V, catalase, ALDH2, medium chain acyl Co-A dehydrogenase, keto-acyl CoA dehydrogenase, and others, leading to decreased ETC, anti-oxidant defense and urea cycle in aged animals [@bib197].
In addition to liver injury, AA-AGE but not NEL stimulated death of neuronal cells [@bib198], suggesting an important role of AA-AGE in neuronal cell death. Consistent with this report, elevated amounts of AGE-albumin adduct, produced by activated microglial cells, were found in alcohol-exposed rat brains and post-mortem brains from human alcoholic individuals compared to those of healthy normal counterparts [@bib199]. Increased oxidative stress can produce AGE-albumin adduct, which elevates the level of RAGE and activates JNK- and p38K-mediated death of neuronal cells. These results are in agreement with the increased expression of RAGE in the brains of alcohol-exposed rats and humans [@bib200]. Consistently, the elevated amounts of AGE-albumin, which was confirmed by mass-spectral analysis [@bib201], were found in the brains of Alzheimer's disease patients than those of the healthy individuals [@bib202]. Inhibition of AGE-albumin production or blockade of RAGE by a chemical antagonist or a specific antibody to RAGE efficiently prevented alcohol-induced brain damage [@bib199], suggesting an important role of AGE-protein adducts in promoting neurodegeneration. Based on these recent results [@bib199], [@bib200], [@bib201], [@bib202], it would be of interest to study the role of AGE-protein adducts in promoting organ damage in the peripheral tissues such as the liver and heart.
Translational research with mitochondrial-targeted antioxidants {#s0040}
===============================================================
As described above, the increased nitroxidative stress in AFLD and NAFLD or after exposure to hepatotoxic agents can stimulate multiple PTMs, including the hydroxyethyl-adducts observed in alcohol-exposed rats [@bib203] and individuals [@bib204] ([Fig. 3](#f0015){ref-type="fig"}). All these PTMs of cellular proteins are likely to contribute to increased ER stress and mitochondrial dysfunction, leading to energy depletion, fat accumulation, altered metabolism, inflammation and necrotic/apoptotic tissue damage ([Fig. 3](#f0015){ref-type="fig"}). Since elevated nitroxidative stress is a critical contributing factor for acute and chronic liver diseases, beneficial effects of many anti-oxidants from natural and synthetic origins have been tested in experimental model systems. These anti-oxidant agents include: [l]{.smallcaps}-arginine, small molecule metabolites (e.g., GSH-ethyl ester and NAC), natural antioxidants \[e.g., vitamins C, E, co-enzyme Q10, alpha-lipoic acid, fish oil containing n-3 fatty acids, betaine and *S*-adenosyl-methionine (SAMe)\], and herbal molecules and polyphenols (curcumin, esculetin, sulforaphane, resveratrol, epigallocatechin-3-gallate, caffeic acid phenethyl ester, and many others) [@bib205], [@bib206], [@bib207], [@bib208], [@bib209], [@bib210], [@bib211]. Some of these antioxidants, contained in many fruits and vegetables, show beneficial effects on obesity, NAFLD and liver injury [@bib205], [@bib206], [@bib207], [@bib208] while other antioxidants are potentially neuroprotective. Recent reports showed that naturally occurring antioxidants such as SAMe and betaine preserved mitochondrial function and proteome in the animal models of AFLD [@bib212], [@bib213], [@bib214] as well as NAFLD [@bib215], [@bib216], [@bib217], partly through blockade of the ROS/RNS production with restoring the physiological levels of SAMe, *S*-adenosyl-homocysteine and GSH. Nonetheless, the list of these antioxidants with therapeutic potential is exponentially growing and will reduce the oxidative burden in various pathophysiological conditions including fatty liver disease.
One of the main problems with some of the natural antioxidants is to guarantee good manufacturing quality control \[e.g., a constant amount of the active component(s) with acceptable levels of impure materials\]. Other problems can be low solubility, poor bioavailability and little mitochondrial transport [@bib208], [@bib209], [@bib210], [@bib211]. To overcome these problems, many compounds with better mitochondrial targeting properties have been synthesized to prevent or treat mitochondrial dysfunction in many pathological conditions [@bib211], [@bib218], [@bib219], [@bib220], [@bib221], [@bib222], [@bib223], [@bib224], [@bib225]. For instance, to accomplish better mitochondrial transport, Szeto and Schiller developed a series of small, cell-permeable, mitochondria targeted antioxidant peptides (Szeto--Schiller or SS-tetrapeptide) that protect mitochondria from oxidative damage [@bib220]. These peptide antioxidants represent a novel approach with targeted delivery of antioxidants to the inner mitochondrial membrane. The structural motif of these SS-peptides centers on alternating aromatic residues and basic amino acids (aromatic-cationic peptides) that can scavenge hydrogen peroxide and peroxynitrite while they inhibit lipid peroxidation. Their antioxidant activities can be attributed to the tyrosine or dimethyltyrosine residue with inhibition of mitochondrial permeability transition and cytochrome c release, thus preventing oxidant-induced cell death. Because these peptides can accumulate \>1000-fold in the inner mitochondrial membrane, they prevent oxidative cell death with EC~50~ in the \~nM range. Similarly, the analogs of SOD-catalase mimetics also showed beneficial effects on many tissues including the liver and brain [@bib221].
In contrast to the SS-peptides or peptide-mimetics, a small molecule triphenyl phosphonium (TPP^+^, a cell permeable lipophilic cation) has been developed to conjugate with various drugs and antioxidants to enhance the transport of target molecules to reduce oxidative stress and damage, as reviewed [@bib222], [@bib223]. The results with mitochondria-targeted ubiquinone (MitoQ) or mitochondria-targeted carboxy-proxyl (Mito-CP) so far have shown promising results in preventing mitochondrial abnormalities and nitroxidative tissue injury in various disease models including Friedreich Ataxia fibroblasts [@bib224], diabetic nephropathy [@bib225], cisplatin-induced renal toxicity [@bib226] and cocaine-induced cardiac dysfunction [@bib227]. In addition, the protective effects of MitoQ and Mito-CP were clearly demonstrated in a mouse model of hepatic I/R injury [@bib228]. In a dose-dependent manner, these mitochondria-targeted compounds blocked the early and delayed nitroxidative stress responses (e.g., HNE/carbonyl adducts, malondialdehyde, 8-OHdG, and 3-nitrotyrosine formation), mitochondrial dysfunction and histopathological signs of liver injury as well as delayed inflammatory cell infiltration and cell death [@bib228]. Consistently, mitoQ was shown to be protective against micro- and macro-vesicular steatosis in AFLD [@bib229] without affecting the levels of CYP2E1 and ALDH2 as well as ethanol-induced mitochondrial respiratory abnormality. The effects of mitoQ were rather mediated through the suppression of ROS/RNS production and their down-stream targets such as protein nitration and HIF-1α stabilization [@bib229]. These promising results from different laboratories suggest that mitochondria-targeted antioxidants are more effective against elevated nitroxidative stress than untargeted natural antioxidants. Because of the recent development of these antioxidant agents and clinical testing [@bib230], [@bib231], we expect approval of some of these beneficial agents in treating mitochondrial dysfunction-related organ damage and hence preventing the health deterioration.
Conclusion and future research opportunities {#s0045}
============================================
We have thus far briefly described five major types of PTMs that promote mitochondrial dysfunction, fat accumulation, inflammation and hepatic injury in alcoholic and nonalcoholic substances ([Fig. 1](#f0005){ref-type="fig"}). Although several PTMs in many mitochondrial proteins were identified in different disease states or under increased nitroxidative stress, it is quite possible that many proteins can be modified by multiple PTMs simultaneously. Therefore, it seems difficult to clearly demonstrate the deleterious effect of each single PTM and which PTM plays a critical role in causing mitochondrial dysfunction and tissue damage. In addition, it is challenging to identify the modified amino acid(s) and their roles in activity changes. It is also still unknown what other proteins, especially those expressed in low amounts in a particular tissue, are modified by different PTMs following exposure to a potentially toxic agent. Furthermore, mitochondrial functions can be altered by other types of PTM such as methylation, O-linked glycosylation (Ser or Thr), N-linked glycosylation (Asn), sumoylation, and ubiquitin conjugation. The identities of the proteins modified by these PTMs and their roles in mitochondrial dysfunction, fat accumulation and tissue injury need to be studied in the future. Finally, the patterns of PTM and their functional roles in other extra-hepatic tissues such as brain, heart, pancreas, adipose tissue and intestine need to be investigated. For instance, binge alcohol exposure [@bib232], [@bib233], [@bib234], [@bib235], [@bib236], [@bib237], [@bib238], [@bib239] and nonalcoholic substances including fructose [@bib240], [@bib241], [@bib242] are known to promote gut leakiness, contributing to more severe inflammatory liver disease. In the case of binge alcohol-induced gut leakiness, CYP2E1-mediated increased oxidative stress plays an important role, since treatment with an antioxidant NAC or a CYP2E1 inhibitor chlormethiazole significantly prevented gut leakiness and liver injury. In addition, Cyp2e1-null mice were resistant to alcohol-mediated gut leakiness, which was clearly observed in the corresponding wild-type mice [@bib238]. It is of interest whether gut leakiness caused by nonalcoholic substances also depends on CYP2E1-mediated oxidative stress. Despite numerous reports on gut leakiness caused by alcohol and fructose [@bib232], [@bib233], [@bib234], [@bib235], [@bib236], [@bib237], [@bib238], [@bib239], [@bib240], [@bib241], [@bib242], the functional roles of redox-regulated PTMs on many intestinal proteins are still poorly understood. Therefore, characterization of the modified proteins and their functional consequences in intestinal epithelial tissues and other tissues are not only good for basic mechanistic studies in understanding tissue--tissue interactions but also provide future translational and clinical research opportunities.
This research was supported by the Intramural Program Fund at the National Institute on Alcohol Abuse and Alcoholism. The authors thank Dr. Klaus Gawrisch for his support. The authors do not have any conflict of interest.
{#f0005}
![Multiple PTM of ALDH2 protein, contributing to its inactivation. Different types of PTM on ALDH2 are illustrated. Most PTMs led to the suppression of ALDH2 activity, although phosphorylation of ALDH2 by the PKCε isoform was shown to stimulate the ALDH2 activity, as recently reviewed [@bib42]. The suppressed activities of ALDH2 and other isozymes, all of which share the 100%-conserved active site Cys residue, likely contribute to increased levels of toxic and carcinogenic acetaldehyde (AcAH) and lipid peroxides (MDA and 4-HNE) following exposure to ethanol, high fat and other toxic compounds. ER stress, mitochondrial dysfunction and tissue injury can be observed in acute and sub-chronic cases. Long-term chronic suppression of ALDH2 and other isozymes can also contribute to fibrosis, cirrhosis and carcinogenesis. The negative sign represents the inactivation of ALDH2 and other isozymes.](gr2){#f0010}
![Overlapping PTMs between AFLD and NAFLD. Overlapping PTMs commonly observed in both AFLD and NAFLD are illustrated. However, acetaldehyde--protein adduct (AA--adduct) [@bib171], [@bib181], NEL-adduct [@bib193], AA-AGE-adduct [@bib193], and hydroxyethyl-adduct [@bib203], [@bib204] seem uniquely observed in AFLD. These PTMs observed in AFLD and NAFLD are likely to suppress the functions of the target proteins, contributing to altered cell signaling, ER stress, mitochondrial dysfunction, fat accumulation and inflammatory tissue injury.](gr3){#f0015}
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Rationale
---------
Early life experience is a determinant of social inequalities in physical, mental, and social well-being \[[@ref1]-[@ref6]\]. Children living in poverty are more likely to suffer from developmental and health problems, and childhood interventions can decrease incidence and prevalence of these problems \[[@ref7]-[@ref12]\]. The proposed project is aimed to assess and improve a childhood intervention: the *Naître et grandir* (N&G) website and newsletter on child development, education, health, and well-being, which includes the Information Assessment Method (IAM) that allows N&G readers to continuously assess, and subsequently the N&G editors to improve, the content shared on the N&G website and newsletter.
Parents with low education and low income, hereafter referred to as parents with low socioeconomic status (SES), typically have a low literacy level (limited ability to acquire, understand, evaluate, and use written information). Low parental literacy level is particularly detrimental to child health: a low literacy level is associated with worse health status, difficulties accessing health care, and poorer preventive health behavior and self-management of health problems \[[@ref13]-[@ref23]\].
Education and income are the most important SES indicators, and together, they are strongly associated with child health status \[[@ref24],[@ref25]\]. According to research on information-seeking behavior, parents with low SES have greater information needs than parents with high SES \[[@ref26]\]. The use of high-quality online information can improve quality of life and have positive family, economic, and social impacts on low-SES parents, including refugee and homeless parents \[[@ref27]-[@ref31]\]. High-quality information and literacy-related interventions can reduce unnecessary calls and visits to health professionals, increase knowledge and self-efficacy, and improve health \[[@ref16]-[@ref19],[@ref32]-[@ref47]\].
N&G \[[@ref48]\] is a website independent from industry funding that provides high-quality information (based on research syntheses and validated by experts) on child development, education, health, and well-being. N&G users can access hundreds of information pages (1 page per topic) that are organized in age group categories, ranging from pregnancy to the age of 8 years. N&G also produces a weekly newsletter to support parents, including those with a low literacy level, having children under the age of 8 years.
N&G partnered with investigators from McGill University to validate and implement the IAM to continuously assess and improve content shared on the N&G website and newsletter \[[@ref49]-[@ref53]\]. In line with the Canadian Institutes of Health Research (CIHR)'s definitions \[[@ref54]\], the IAM is a knowledge translation tool for monitoring N&G information use, and its impact on parents is measured by expected health/well-being benefits. It is also fostering parent engagement by enabling parents to interact with N&G editors by providing feedback on the information content. The IAM questionnaire includes 7 questions (with clickable answers and 2 comment boxes), allowing users to rate on the situational relevance, cognitive/affective impact, intention to use and expected benefits of specific information content (N&G information page), and write comments.
During our pilot phase (September 1, 2014, to August 31, 2016), we collected 34,021 IAM ratings (completed IAM questionnaires) from parents, relatives, and professionals (education, health, and social services) who read N&G content. In line with studies on social inequalities in Web information use \[[@ref55]-[@ref61]\], the statistical analysis of these ratings revealed a social gradient: low-SES parents underuse N&G and the IAM ([Figure 1](#figure1){ref-type="fig"}). Results also indicated that low-SES parents are more likely to report decreased worries and increased confidence as a result of using N&G information \[[@ref50]\]. There is a need to understand this gradient to improve N&G content and reach low-SES parents and a need to explore how the use of knowledge translates into health and well-being outcomes for low-SES parents and children.
{#figure1}
Goal and Objectives
-------------------
This study (research protocol) has been recently funded by CIHR. As per CIHR's definition of knowledge translation ("making people aware of knowledge and facilitating knowledge use to improve health") \[[@ref62]\], our goal is to improve how low-SES parents engage with and use online information about child development, education, health, and well-being; learn how this format for knowledge sharing translates to improved health and well-being outcomes for low-SES parents and their children; and explore their interaction with Web editors to influence content (responsible for websites' content). On the basis of the N&G team\'s (the director and 2 coordinators) questions and using the Knowledge-to-Action framework \[[@ref63]\], we determined 4 research objectives:
- Objective 1: Identify low-SES parents' views on barriers and facilitators to accessing, understanding, and using N&G information content and to interacting with N&G editors via the IAM.
- Objective 2: Adapt N&G content/functions and the IAM tool (user-centered design) to reduce barriers and optimize facilitators identified in phase 1 and implement new versions labeled N&G+ and IAM+.
- Objective 3: Evaluate whether N&G+ and IAM+ result in a higher proportion of low-SES parents engaging with the N&G site and the content provided and interacting with editors via the IAM.
- Objective 4: Describe how expected health and well-being benefits of N&G+ information use (reported via IAM+) are experienced by parents with low SES and their children.
Systematic Literature Review and Theoretical Model
--------------------------------------------------
The nominated principal investigator (NPI) led a CIHR-funded systematic review and proposed a comprehensive harmonized typology of (1) outcomes associated with the use of online consumer health information and (2) conditions (network and resources) leading to these outcomes \[[@ref64]-[@ref67]\]. A total of 68 studies were analyzed using framework synthesis \[[@ref68],[@ref69]\]. This synthesis was based on an initial framework derived from information studies \[[@ref49],[@ref70]-[@ref78]\] and led to propose an innovative theoretical model ([Figure 2](#figure2){ref-type="fig"}) \[[@ref66],[@ref67]\]. This model includes positive and negative health outcomes of online consumer health information in a primary care context and is (information) consumer oriented.
The model comprises 13 main concepts (42 factors and outcomes) and will be used in this study to inform data collection and analysis, for example, the phase 1 qualitative interview guide and interpretive thematic analysis. The model will contribute to establish a chain of evidence linking ultimate outcomes of information (such as health outcomes), intermediary outcomes (such as cognitive impact of information), and conditions associated with outcomes (information needs and seeking behaviors and contextual factors).
{#figure2}
### Outcomes
The model includes an organizational level of information outcome (eg, increased or decreased use of health services) and 4 individual levels of outcomes of information. The latter outcomes reflect how information is valuable from the consumer's perspective: situational relevance, cognitive impact, use of information (conceptual, legitimating, instrumental, and symbolic use), and subsequent health and well-being outcomes.
### Conditions
The model includes conditions associated with outcomes of information in relation to a specific situation: a particular information object is acquired or delivered (eg, a Web page or a newsletter) in a particular situation (eg, before or during or after an encounter with someone) directly or with help from a relative or a professional. The main conditions are information needs and seeking behaviors; individual characteristics (such as electronic health literacy); social and technical factors (such as social networks); relationships with professionals (such as teachers, clinicians, and social workers); and access to education, health, and social services. The first phase of this study will specifically look into conditions experienced by low-SES parents, such as media competence.
Direct acquisition of online information depends on an impetus to search (motivation), momentary internet connection, internet search skills, and one's ability to understand the content that may or may not be adapted to their individual literacy level \[[@ref79]-[@ref81]\]. Approximately 95% of parents of preschool children have direct individual access to the internet in Quebec \[[@ref61]\], and parents can also access it at office workplaces and public libraries \[[@ref79]-[@ref81]\]. Literacy level is generally defined as the degree to which a person has the ability to "acquire, understand, evaluate, and use information" needed to obtain services and make appropriate decisions \[[@ref23]\]. Computer literacy, information literacy, and health literacy are interdependent (eg, a person with a low literacy level has a low level of health literacy). Culture is central in literacy, and one's literacy level depends on one's ability to understand systems of symbols from one's own culture or a dominant culture and language, for example, immigrants and refugees may have a higher literacy level in their country of origin compared with their adoptive country \[[@ref23]\]. As mentioned in our model, the literacy level is *situational* and *contextual*, given that a social network can compensate for an individual's low literacy level.
Mediated (by someone else) acquisition of information is very common. The absence of an individual connection 24/7 is no longer the primary barrier to seeking health information \[[@ref80]\]; for example, our pilot data show that 18.1% of IAM ratings concerned N&G information seeking for someone else's child (eg, relative's, friend's, neighbor's, or client's child). Even homeless parents or recent immigrants and refugees can acquire online consumer health information directly or mediated by their social network, including community organizations; public libraries; and education, health, and social services \[[@ref23],[@ref27]-[@ref31]\]. Indeed, information studies show that consumers combine mediated information with direct acquisition of online health information, the latter allowing them to probe information provided by professionals \[[@ref64]-[@ref67]\]. For analytical purpose, we conceive these combinations as the interpenetration of social systems centered on communicative action, for example, the health system (mediated access) and a consumer parent system (direct access and mediated access via their social network) \[[@ref82]-[@ref85]\].
Significance
------------
Our project focuses on all Canadian parents with low SES who acquire (direct or mediated) online information about child development, education, health, and well-being. In Canada, 42% of nonelderly adults \[[@ref16],[@ref17]\] and 15% of children live in low-income households \[[@ref86]\]. In Quebec, 36% of parents report living in poverty \[[@ref87]\]. The demand for information is very high, resulting in frequent internet searches; for example, parents search the internet for child-related information on average 1.3 times per week in Quebec, and 73% of parents report that the internet constitutes their first-line source of parenting information \[[@ref88]\].
######
Main features of *Naître et grandir* and *AboutKidsHealth*.
Main features *Naître et grandir* \[[@ref48]\] *AboutKidsHealth* \[[@ref94]\]
------------------------ -------------------------------------------------------------------------------------------------- -------------------------------------------------------------
Website annual traffic 24 million worldwide visits 16.8 million worldwide visits
Information Developmental, educational, health, and well-being information about children aged under 8 years Health information about children aged under 19 years
Language French English, French, and 10 other languages
Targeted audience Parents, relatives, and caregivers Parents, relatives, and caregivers and health professionals
Specifically, our project is important for about half the population of parents of children under 19 years, namely, parents with a low literacy level. Low literacy is a major concern in Canada \[[@ref16],[@ref17],[@ref19],[@ref20],[@ref23]\]. Results of the 2011 to 2012 survey of a representative sample of 25,267 Canadian parents aged 16 to 65 years showed that 49% of parents have a low literacy level: \[[@ref89],[@ref90]\] they have difficulty finding, understanding, and using information presented in a dense or lengthy text; navigating complex digital texts; interpreting and evaluating information (constructing meaning); and disregarding irrelevant or inappropriate content when there is competing information (including when correct content is more prominent). Furthermore, our results may help any parent who faces a transitory low literacy level because of a stressful situation. This interdependence between information and emotion is well established in the information literature \[[@ref91],[@ref92]\].
Our project nevertheless does not focus on the few *information-poor* parents without access (direct or mediated) to online information. In the United States, only 1% of 18- to 29-year olds did not use the internet in 2015 \[[@ref93]\]. The Quebec 2015 survey of a representative sample of 23,693 parents of preschool children showed that only 1.5% of parents never know where to find child information \[[@ref61]\]. According to information studies, the *information-poor* parents (1) perceive themselves as persons who cannot be helped, (2) adopt self- or group-protective behaviors, (3) are secretive and mistrust others, and (4) consider exposure to information as a risk (harm outweighing benefits) \[[@ref30]\].
Our project engages a partnership between N&G and *AboutKidsHealth* (AKH) \[[@ref94]\], both popular comprehensive Web-based resources ([Table 1](#table1){ref-type="table"}). Both resources contain text with a readability level below high school (this level may be higher for disease-related information), visual illustrations, and complementary simple illustrative videos. For each Web page, an audio help system highlights words as they are spoken. In 2016, our librarian did an environmental scan of Canadian parental websites and found multiple specialized resources but only 12 comprehensive resources (including N&G and AKH).
Although this study is a priority in a Canadian context, our results can be important to and applied by all Web editors who provide health information, thereby benefitting all adults with a low literacy level (not only parents). For example, Health Canada policies recommend funding projects that aim to improve access to, and conduct research on, information for Canadians with a low literacy level \[[@ref23]\]. Information systems are key components of health systems and crucial to meet human rights and the democratic right to know and communicate \[[@ref31],[@ref95]-[@ref100]\]. Among adults, the most frequent searches for information are about health \[[@ref101]\]. The demand for high-quality low-literacy health information is, hence, very high.
Methods
=======
Organizational Participatory Research Approach
----------------------------------------------
Our project uses an organizational participatory research (OPR) approach. OPR is a form of integrated knowledge translation that blends action research and organizational learning to undertake research with organizations and improve practice \[[@ref102]-[@ref107]\]. N&G (organization) proposed the research questions and will participate in all research steps. The NPI has OPR experience and expertise. The NPI led a CIHR-funded systematic review on OPR key processes and outcomes \[[@ref107]\]. A steering committee composed of the NPI, the coprincipal investigator, the partner (principal knowledge user), another knowledge user (AKH), and 2 low-SES parents will meet before and after each phase for planning and interpreting results, respectively. All team members will be consulted and have the opportunity to influence the steering committee's decision making.
Methodology and Methods
-----------------------
A mixed-methods multiphase design will be used \[[@ref108],[@ref109]\]. Ethical approval has been recently obtained from the institutional review board (IRB) of the Faculty of Medicine at McGill University (IRB\#A10-E69-17B). Phases 1, 3, and 4 will be informed by our theoretical model ([Figure 2](#figure2){ref-type="fig"}).
### Phase 1 (Objective 1): Identify Barriers and Facilitators to Use Naître et Grandir and Interact With Editors Via Information Assessment Method
#### Design
We will conduct an exploratory qualitative interpretive study to have a better understanding of barriers and facilitators in using N&G and interacting with editors via IAM \[[@ref110]\]. Qualitative research is appropriate as it provides in-depth descriptions from the stakeholders' viewpoint and helps researchers to understand their constitutive elements and variants \[[@ref111]\].
#### Setting and Participants
Participants will be 20 users and 20 nonusers of IAM and N&G (additional participants will be recruited to achieve data saturation). First, users will be recruited among low-SES parents who indicate that they agree to be contacted when they complete an IAM questionnaire on N&G and who have used N&G and IAM at least once. N&G will send the group an invitation to participate in a research project via an email containing the research team's full identification and contact information. The research team will only communicate with N&G users who responded with an interest to participate in the research project via email or phone. Eligible parents will have no high school diploma and have an annual family income lower than Can \$40,000 (Quebec poverty line). Second, nonusers (10 who never used the IAM and 10 who never used N&G) will be recruited by a principal investigator (CL) and a collaborator (GD, N&G Director) who work with community organizations for parents with low SES.
#### Data Collection
A research assistant having experience in qualitative interviews with low-SES persons will conduct individual 60-min semistructured face-to-face interviews at the location of each participant's choice. In line with sociological and information studies \[[@ref112],[@ref113]\], interviews will explore how participants experience, conceptualize, perceive, and understand aspects of N&G information and the IAM. The interview guide will consist of open-ended questions addressing the participants' routines in terms of internet use (both for their personal and children's development information needs), their experience in the use of IAM and N&G's information, the facilitating aspects of N&G and IAM, perceived barriers in their experience, and suggestions to improve the N&G website and IAM. On the basis of the theoretical model, the interview guide will be developed in a simplified language with input from team members and the steering committee. We will conduct 2 pilot interviews to ensure the questions are comprehensible for the participants. The recruited respondents will receive Can \$20 as compensation for their time. In addition to the research assistant's observation notes, interviews will be audio recorded and transcribed verbatim.
#### Data Analysis
Two principal investigators (PP and CL) have experience in qualitative research and 1 (CL) has experience in qualitative research with populations in situations of vulnerability. With the research assistant and a research trainee, they will read notes and interviews and meet regularly to build memos (meeting minutes with arguments for and against each analytical decision, as rigor is mainly based on researchers' reflexivity in this design \[[@ref110]\]), case summaries, and themes (definitions and key examples) using hybrid deductive-inductive thematic analysis \[[@ref114],[@ref115]\]. The research assistant will manage the analysis using specialized software (NVivo11). Themes will be derived from our theoretical model ([Figure 2](#figure2){ref-type="fig"}; deductive coding) and emerge from the data (inductive coding). Major themes will be barriers and facilitators. For each type of participant (user/nonuser), additional parents will be recruited up to saturation (no new parents' views of barrier or facilitator). Then, we will harmonize themes \[[@ref116]\]. For each theme and definition, we will identify terms, confirm the usage of these terms in reference documents, distinguish correct from incorrect usage, and retain terms that facilitate unambiguous communication. This will lead us to classify, group, and clarify barriers and facilitators in a coherent taxonomy, which will be reviewed by team members and the steering committee. Findings will provide recommendations for improving IAM and N&G.
### Phase 2 (Objective 2): Improve Naître et Grandir and Information Assessment Method
Two team members (FL and GD) are head directors of the N&G website and will lead this phase. Phase 1 findings will inform the production and implementation of IAM+N&G+ (user-centered design). This version will integrate the perspectives of low-SES parents, the steering committee, research team members, and N&G staff on how to overcome barriers and optimize facilitators. Production of this version will be planned with 2 Web editors and 2 Web engineers from the N&G team. According to Web engineers, 6 months is an ample time frame for production and beta testing. All necessary resources at N&G will be made available to implement IAM+N&G+. Two half-day meetings with principal investigators and N&G staff have been deemed sufficient to plan changes with editors and engineers. The N&G Director stated that she will implement what will be requested by the phase 1 participants, as her mandate (N&G mission) is to specifically address information needs of parents with low SES. For their part, AKH will redesign their website, integrate phase 1 results in this study, and implement IAM+.
### Phase 3 (Objective 3): Evaluate Use of Information Assessment Method+ Naître et Grandir+, Information Use, and Benefits
#### Design
We will conduct a quantitative prospective longitudinal study to evaluate the impact of the intervention (IAM+N&G+).
#### Setting
Weekly numbers of N&G sessions and IAM ratings will be monitored over 2 years. A session starts when someone opens a Web page and ends when the person (or a relative with the same IP address) does not use the website for more than 30 min. For each quintile of SES, data will be collected for 9 months preimplementation (IAM/N&G) and 9 months postimplementation (IAM+N&G+). To avoid bias related to the *novelty effect*, data collected in the 3 months immediately following the implementation will be excluded from the analysis. The chosen periods also ensure seasonal comparability of the collected outcome data (before and after intervention).
#### Participants
All N&G readers and IAM raters across Canada will participate.
#### Measurement
N&G and McGill already use Google Analytics, an objective and reliable automatic data collection of N&G readers' demographic characteristics and website use behavior (based on javascript codes in each Web page) \[[@ref117]-[@ref120]\]. IAM raters' demographic data are collected with a questionnaire (linked to an anonymized identifier). The validated IAM questionnaire \[[@ref53]\] collects self-reported information use and subsequent expected health/well-being benefits for parents and their child.
######
Types of outcomes monitored weekly.
Type Outcome Description and data source
------ ------------------- -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
A1 N&G^a^ sessions Weekly number of unique sessions (Google Analytics)
A2 N&G read Weekly proportion of sessions with at least 1 page entirely read (Google Analytics)
A3 IAM^b^ ratings Weekly number of IAM questionnaires submitted
B1 N&G mediated Weekly proportion of IAM ratings information used for the child of someone else
B2 N&G used Weekly proportion of IAM ratings information used for oneself and one's child
C Expected benefits Weekly proportion of IAM ratings expected health or well-being benefit for a parent and child, including at least 1 of the following self-perceived benefits (IAM choices of response): improvement of the health or well-being of a child, being less worried, prevention of a problem or the worsening of a problem, handling a problem, and being more confident to decide something with someone else
^a^N&G: *Naître et grandir*.
^b^IAM: Information Assessment Method.
#### Outcomes
In line with CIHR knowledge translation guidance \[[@ref54]\], 3 types of outcomes will be considered weekly ([Table 2](#table2){ref-type="table"}): (A) N&G and IAM use, (B) self-reported N&G information use, and (C) subsequent expected benefits.
#### Covariates
The main predictor will be participants' SES as determined by the Quebec Index of Material and Social Deprivation, a validated ecological measure of the (education, income, and employment) disadvantage of a given geographic area (postal code) \[[@ref121],[@ref122]\]. For each session, an index will be automatically assigned using (1) the reported postal code and the Canadian Deprivation Index Assignment Program (CDIAP) of the Public Health Agency of Canada when the data source is IAM and (2) the postal code matched to the session geotag via the "Dissemination Area Boundary File" of Statistics Canada and the CDIAP when the data source is Google Analytics. This will allow for the identification of low-SES participants (highest quintile of material and social deprivation). Using Google Analytics, participants' demographic variables that will be considered in all analyses are age, gender, rural/urban location, and province. In addition, Web page--specific variables will be included to assess potential confounding and effect modifications. For each session, the type of device (Google Analytics: phone, tablet or laptop, and desktop), the audio-guide use (Google Analytics: yes or no), and the readability score of pages entirely read (automatic extraction and measurement using a text classifier validated for French) \[[@ref123]\] will be collected.
#### Study Size
On the basis of our 2-year pilot data (2014/09-2016/08), we anticipate about 5 million N&G sessions and 15,000 IAM ratings from Canadian participants during each period (pre- and postimplementation).
#### Anticipated Results
The IAM+N&G+ will result in an increase of all outcomes (N&G and IAM use, self-reported N&G information use, and subsequent expected health and well-being benefits for parents and children such as *decreased worries* and *health improvement*, respectively).
#### Statistical Analysis
Linear mixed modeling will integrate spatial analytics (geomatics) and account for the clustered nature of the data. The pre/post status, the SES quintile of deprivation, and potentially confounding variables will be included as fixed-effects covariates. The inference model will incorporate random effect terms for individual variables and Web page--specific variables. For each outcome, an interaction term of the SES quintile of deprivation and the pre/post status will be included to assess the pre/post change for each quintile. Estimated regression coefficients and variance parameters will be reported along with appropriate confidence intervals.
### Phase 4 (Objective 4): Describe Parent and Child Outcomes of Naître et Grandir+ Information Use
#### Design
We will conduct a qualitative interpretive study to generate an in-depth description of parent and child outcomes from the parents' perspective \[[@ref124]\]. On the basis of our systematic review \[[@ref66],[@ref67]\], we anticipate 3 types of parent outcomes (decreased worries, increased confidence, and self-management) and 2 types of child outcomes (prevented problem and improved development, health, and well-being). Our pilot data (34,021 IAM ratings) also suggest parents rarely report potential negative consequences of using information, for example, vaccine adverse effect (n=183; 0.5%). In outcomes research, qualitative methods are appropriate for exploring complex outcomes from the stakeholders' perspective, such as life experiences \[[@ref111],[@ref125]-[@ref131]\]. In our study, information can influence parent decision making, but the relationship between information and decision is not simple. Our qualitative study will (1) identify causal events, (2) map them in a complex causal network, and (3) build a chronological chain of qualitative evidence between N&G information and a parents' decision that affects knowledge, attitude, or behavior \[[@ref131],[@ref132]\].
#### Participants
Participants who received the N&G+ newsletter, used IAM+ at least once, have agreed to be contacted for research, and live in a high deprivation area will be approached for recruitment. The recruitment procedure will be the same as for phase 1. In 2015, more than 24,000 families received, upon request, the weekly newsletter. We will begin by recruiting 30 parents (and continue recruitment, as needed, to achieve data saturation) who report, via IAM+, positive health and well-being outcomes and/or potential negative consequences of information (prioritizing those who live in the areas of highest deprivation). As we are looking for individual stories for about 6 types of outcomes (some less frequently reported than others), data saturation may not be reached with fewer than 50 participants \[[@ref111],[@ref133]-[@ref135]\].
#### Data Collection
The research assistant having experience in qualitative interviews with low-SES persons will conduct 60-min face-to-face interviews to elicit participants' IAM+ ratings. The interview guide will be based on our theoretical model ([Figure 2](#figure2){ref-type="fig"}) and input from team members and the steering committee. To stimulate recall (memory), the interviewee will be given the list of recent texts they rated and their ratings. For each newsletter, the research assistant will ask open questions regarding what happened to them and/or their children that led them to report an expected outcome (positive or negative). Participants will be interviewed twice 3 months apart to increase the number of described outcomes and avoid fatigue of long interviews. Participants will receive Can \$20 per interview as compensation for their time. In addition to the research assistant's observation notes, interviews will be audio recorded and transcribed verbatim.
#### Data Analysis
Two principal investigators (PP and CL) have experience in qualitative research. With the research assistant, a collaborator anthropologist having expertise in life histories (MB) and a research trainee will read notes and interviews and meet regularly. For each case (information used with at least 1 outcome experienced by a parent and a child), they will interpret the data in the form of a *small story* \[[@ref112],[@ref136]-[@ref138]\]. This method allows researchers to describe a person's individual experience (including all perceived influences) and helps the researcher understand the individual's attitude and behavior \[[@ref33]\]. In research meetings, 2 questions will be answered through recorded discussion of arguments for and against each analytical decision (rigor based on sharing reflexivity): is the case story clear? and are the causal network and chain of evidence trustworthy? Disagreements about clarity and trustworthiness of case stories will be resolved with 3 other team members having experience in qualitative research (Bouthillier, Thoër, and Smythe). Case stories will be reviewed by all team members and the steering committee. This may detect issues and result in the principal investigators and research assistant revising parts of their analysis.
#### Anticipated Results
This will generate up to 10 case stories per outcome type, being the first in-depth qualitative description of low-SES parents' perspective on outcomes of online child information use.
Results
=======
The project was funded in 2017 by the CIHR and received an ethics approval by the McGill University's IRB. Data collection for phase 1 was completed in 2018. Phases 2 to 4 will be completed by 2020. Findings from this study will be used to develop a free toolkit, useful to all Web editors, with recommendations for improving health information for low-SES persons and interactions with them using IAM. Results will be published in peer-reviewed journals and presented at national and international scientific conferences.
Discussion
==========
Direct Impact on *Naître et grandir* and *AboutKidsHealth*
----------------------------------------------------------
Any improvement of knowledge translation tools and websites such as IAM, N&G, and AKH can have an important impact as it affects a large population of individuals with a low literacy level. In fact, the effectiveness of childhood education interventions has been demonstrated repeatedly \[[@ref16]-[@ref19],[@ref23],[@ref27]-[@ref47]\]. In our 2-year pilot data, parents expected health and well-being benefits (for themselves or their child) from using N&G information in 65.4% of all IAM ratings (n=34,021) \[[@ref50]\]. In accordance with knowledge translation and implementation research \[[@ref62]\], we will not replicate effectiveness studies and rather focus on improving interventions that work.
Our project can improve engagement (number of visits) with websites providing high-quality low literacy information; in turn, online parenting information improves parents' knowledge, attitudes, and behaviors \[[@ref139]-[@ref145]\]. Given our integrated knowledge translation (participatory research) approach, N&G and AKH will adapt their content for low-SES users as we generate results. In addition, IAM+ will better support low-SES parents' interactions with Web editors and empowerment. Thus, our results can immediately benefit millions of information users with a low literacy level (children's parents and relatives), for example, in Canada ([Table 3](#table3){ref-type="table"}).
IAM+N&G+ can be seen as an innovative intervention that complements traditional literacy programs (eg, family literacy classes). Indeed, interventions that somewhat compensate for a low literacy level can greatly improve parents' and children's health and well-being \[[@ref23]\]. Thus, our results can have a positive impact on 55% of the Canadian working age, those who have a low level of *health literacy* and need compensatory help to manage their health \[[@ref23],[@ref89],[@ref90]\].
Specifically, IAM+N&G+ can help Canadians in francophone minority communities, as 500,000 annual N&G website visits originate from them. Research with these communities is very underfunded, thus a CIHR priority \[[@ref146]\]. As stated in the October 2016 report of the Commissioner of Official Languages \[[@ref147]\], early childhood development in these communities *is hindered by a lack of resources*, and *the absence of specific funding* has left them *vulnerable and often incapable of meeting their needs*.
######
Targeted population in Canada.
Across Canada IAM^a^ *Naître et grandir* (new version) IAM *AboutKidsHealth* (new version)
------------------------------------- ---------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------
Families^b^ 1.3 million families have French as mother tongue (couples with children and single-parent families with at least 1 child) 5.8 million families (couples with children and single-parent families with at least 1 child)---all languages
Children's parents and relatives^b^ 2.8 million adults aged 20 to 69 years have a low literacy level (and French as mother tongue) 12 million adults aged 20 to 69 years have a low literacy level
Website visits from Canada 8.3 million^c^ visits in 2015 mainly from Quebec (7.8 million), Ontario (304,000), and New Brunswick (70,000) 1.5 million visits in 2015 mainly on Web pages in English (58%), French (16%), Spanish (14%), and Arabic (5%)
^a^IAM: Information Assessment Method.
^b^Statistics Canada 2015.
^c^76% of parents of children aged under 8 years consult *Naître et grandir* in Quebec (on average 1.3 times per week).
Potential Impact Beyond *Naître et grandir* and *AboutKidsHealth*
-----------------------------------------------------------------
Results will inform the development of an open access online tool kit on how to adapt websites and IAM+ for a low literacy audience (a growing consideration of many Web editors). The tool kit will be designed to be generalizable to all types of online consumer health information in Canada. It will include guidance with 3 main messages: (1) how to produce simple information in lay language with audio and visual content, (2) how to better interact with low-SES persons using IAM+, and (3) how to use consumers' IAM+ ratings and feedback to continuously optimize information content. The tool kit will be freely available to any Web editor via the Quebec SPOR-SUPPORT Unit and IAM websites. The NPI has experience in developing tools and leads the method development platform of the Quebec SPOR- SUPPORT Unit.
Our primary knowledge translation goal for knowledge user audiences will be to raise awareness about our main messages and drive attention to the tool kit. A secondary knowledge translation goal will be to support the implementation of the toolkit by Web editors. Knowledge users will disseminate our work in multiple Canadian organizations. As per reviews on scaling up \[[@ref148]-[@ref150]\], we will update our 2016 environmental scan and contact all Canadian websites targeting parents. A specialized librarian (collaborator) will reach additional health websites through librarian listservs and peer networks. Team members will raise general awareness for the tool kit and deliver our main messages to a variety of academic and nonacademic knowledge user audiences via a range of knowledge translation strategies (conference presentations, open access peer-reviewed publications, plain language summaries and grey literature, social media, and networks). Multiple knowledge translation channels will be used including Twitter (\@UniteSoutien) and websites such as the Quebec SPOR-SUPPORT Unit and the IAM websites \[[@ref151],[@ref152]\].
Contribution to Scientific Knowledge
------------------------------------
Our project will advance knowledge on the value of online information for, and interaction with, low-SES persons to strengthen health systems. On the one hand, this project is the first to systematically explore health and well-being outcomes for low-SES parents and children associated with parenting information websites. Most studies focus on discussion forums that can be intimidating to low-SES parents \[[@ref140],[@ref142],[@ref143],[@ref153],[@ref154]\]. On the other hand, many studies concern relational marketing and website feedback buttons \[[@ref52],[@ref155]-[@ref158]\], but we know of none that addresses interactions between low-SES persons and Web editors (such as interaction through IAM+).
Timeline and Potential Challenges
---------------------------------
Overall, 3 years will be needed to complete this project ([Figure 3](#figure3){ref-type="fig"}). Our phases are simple and well defined, and our planning is realistic. In phases 1 and 4, we anticipate no recruitment difficulties. Many parents have already agreed to be contacted for research, and N&G will facilitate the recruitment. Phase 2 is feasible, and the deliverables are realistic given that N&G has already implemented IAM, McGill and N&G have a strong OPR partnership, and N&G is committed to implementing IAM+N&G+. In phase 3, the large number of N&G readers guarantee a large sample. To control for potential seasonal variation in the number of sessions pre- and postimplementation, monitoring will be conducted during the same seasons. To control for contextual changes, variables such as type of device, age, gender, and location are covariates. We have opted for a prospective longitudinal study as monitoring is embedded in N&G routines. In addition, although randomization would provide a higher level of evidence, it was considered unethical to randomly assign low-SES parents to N&G+ or N&G information and impractical for such a popular website (contamination bias) \[[@ref159]\]. Moreover, 3 typical potential sources of biases in longitudinal designs may not affect our study: the use of proportions will control for historical events, there will be no *respondent fatigue* as our 2-year pilot data showed that the monthly number of IAM raters is stable, and the social desirability bias will affect pre- and postimplementation periods in a similar manner.
{#figure3}
Considering the commitment, expertise, and networks of our team, we expect no major challenges. This project is highly facilitated by the participatory research approach (a form of integrated knowledge translation). In addition, it is sustainable as our results will be applied and sustained by 2 longstanding organizations. N&G is fully funded since 1998 by the Lucie et André Chagnon Foundation (one of the largest philanthropic agencies in Canada) and is a priority of this foundation. AKH is owned by the Hospital for Sick Children (Toronto) and is fully funded since 2004 by multiple maternal and child health agencies.
Conclusions
-----------
The results of this study will provide a deep understanding of how low-SES parents use online child information and interact with Web editors. Following the implementation of IAM+N&G+, results will also elucidate subsequent health outcomes for low-SES parents and children after interaction with Web editors have been optimized. Thus, our results can immediately benefit millions of information users, children's parents and relatives in particular, with a low literacy level.
The authors wish to thank Dr Suzanne Smythe (Simon Fraser University, Canada), Mr Sean Schurr (AboutKidsHealth, Canada), Maria de Carmen Reyes (Centro de Investigación en Geografía y Geomática, Mexico), and Ms Catherine Chouinard (Avenir d'enfants, Canada) for their collaboration in developing this proposal. N&G is funded by the philanthropic organization "Fondation Lucie et André Chagnon." PP holds an investigator salary award from the Fonds de recherche en santé du Québec. This study has been funded by the CIHR (\#PJT153200).
Conflicts of Interest: None declared.
AKH
: AboutKidsHealth
CDIAP
: Canadian Deprivation Index Assignment Program
CIHR
: Canadian Institutes of Health Research
IAM
: Information Assessment Method
IRB
: institutional review board
N&G
: Naître et grandir
NPI
: nominated principal investigator
OPR
: organizational participatory research
SES
: socioeconomic status
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Classical trace amines (TAs), including tyramine, beta-phenylethylamine (β-PEA), tryptamine and octopamine, have been implicated in a number of neuropsychiatric disorders associated with monoaminergic dysfunction, including schizophrenia, major depression, and Parkinson's disease ([@B9]; [@B70]; [@B11]; [@B15]; [@B4]; [@B85]; [@B59]; [@B45]; [@B5]). TAs are structurally, metabolically and functionally related to monoamines, and are synthesized in nerve terminals by decarboxylation of the amino acids that serve as precursors for dopamine (DA), noradrenaline, and serotonin ([@B4]). TAs are present in mammalian tissues at very low (nanomolar) concentrations ([@B34]), and are stored in monoaminergic nerve terminals where they are released together with monoamines ([@B11]). TAs are recognized as substrates for monoamine transporters, suggesting similarities between the regulation of extracellular levels of TAs and monoamines ([@B58]; [@B16]; [@B82]; [@B54]; [@B63]). Neuroanatomical observations and cellular studies indicate that TAs have a modulatory influence on monoaminergic neurotransmission, in particular on dopaminergic transmission, which is expressed across multiple cerebral structures ([@B31]; [@B95], [@B96]; [@B86]; [@B45]; [@B61]). A reduction in TA levels has been proposed to be associated with depressed states ([@B77]; [@B79]; [@B21]; [@B76]; [@B91]; [@B11]). TA levels are enhanced by inhibition of monoamine oxidase A and B in animals where the corresponding genes have been deleted ([@B41]).
For a long time the pharmacological effects of TAs were attributed to a direct interference with aminergic pathways, up until the cloning and characterization of a large family of G protein-coupled receptors, named trace amine-associated receptors (TAARs) which were found to be activated by TAs ([@B8]; [@B14]). These receptors responded to the endogenous TAs along with several amphetamines. Outside the central nervous system (CNS), TAAR1 is expressed in pancreatic β-cells, stomach, intestines, thyroid gland, and leukocytes ([@B92]; [@B45]; [@B5]). It is therefore interesting that endogenous 3-iodothyronamine (T~1~AM), which is a derivative of thyroid hormone (thyroxine, or T~4~), has been found to be an endogenous agonist at TAAR1 ([@B80]; [@B36]; [@B26]). The reduction of core temperature and cardiac output induced by T~1~AM, which contrast to the effects induced by thyroxine itself, have been suggested to be mediated by TAAR1 activation ([@B80]; [@B103]; [@B18]; [@B26]; [@B33]; [@B22]). In the brain, TAAR1 is enriched in the major nuclei of the monoaminergic system such as the ventral tegmental area (VTA), substantia nigra pars compacta (SNc), locus coeruleus and raphe nuclei as well as their projection targets, the hypothalamus, layer V pyramidal neurons of prefrontal cortex (PFC), caudate nucleus, putamen, nucleus accumbens (NAc), hippocampus, and amygdala ([@B8]; [@B14]; [@B94]; [@B55]; [@B29]; [@B45]; [@B68]; [@B5]). However, the highest TAAR1 mRNA levels are clearly found in the dopaminergic cell groups (VTA and SNc) as compared to other brain regions ([@B57]). The detailed expression pattern of TAAR1 among the different neuronal populations has not been yet fully defined. Nonetheless, since its discovery, TAAR1 has emerged as a modulator of monoaminergic functions and a mediator of psychostimulant effects ([@B63]; [@B97]; [@B22]).
Trace amine-associated receptor 1 is coupled with stimulatory G~s~ proteins, but its signaling also involves the G protein independent β-arrestin2/Akt/Glucogen Synthase Kinase-3β (GSK-3β) pathway ([@B35]). The latter pathway is known to be downstream of D~2~ receptors ([@B30]; [@B68]). There is evidence that TAAR1 interacts directly with D~2~ receptor by forming heterodimers, however, a peculiar aspect of the receptor is its intracellular residence ([@B68]). This intracellular localisation of the receptor has been indicated by experiments with tagged TAAR1 proteins where it was observed that the chimeric molecules showed robust intracellular distribution ([@B14]; [@B98]; [@B35]).
The elucidation of TAAR1 function has been greatly facilitated by the development of selective pharmacological tools and the generation of mutant TAAR1 animal models. *N*-(3-ethoxy-phenyl)-4-pyrrolidin-1-yl-3-trifluoromethyl-benzamide (EPPTB) is a selective TAAR1 antagonist ([@B10]; [@B87]), whereas several agonists, including RO5166017, binds to TAAR1 with high affinity ([@B73]). TAAR1 knockout (KO) mouse lines have been generated to further delineate the role of TAAR1 ([@B94]; [@B55]; [@B22]). There are no gross behavioral abnormalities in TAAR1 KO mice, but upon closer examination they show an impulsive and perseverative phenotype ([@B94]; [@B28]). Baseline extracellular DA levels in striatum are similar between wild-type (WT) and TAAR1 KO mice ([@B55]; [@B22]; [@B48]; [@B68]). However, electrophysiological experiments have shown that dopamine neurons in VTA and serotonin neurons in dorsal raphe nucleus (DRN) from TAAR1 KO mice display increased firing rates compared with WT mice ([@B55]; [@B10]; [@B62]). The endogenous TA, tyramine, specifically decreased the spike frequency of VTA neurons in WT but not in TAAR1 KO mice ([@B55]).
Trace amine-associated receptor 1 KO mice have repeatedly been shown to display increased sensitivity to amphetamines, measured as an enhanced increase in locomotor activity and enhanced striatal release of DA compared with WT animals ([@B94]; [@B55]; [@B62]; [@B1]). Accordingly, TAAR1 is thought to act in the mesocorticolimbic system to regulate cocaine-seeking behavior ([@B57]). We have also found an increased behavioral responsivity toward [L]{.smallcaps}-dihydroxyphenylalanine ([L]{.smallcaps}-DOPA) in TAAR1 KO mice rendered unilaterally dopamine denervated by 6-hydroxydopamine injections in the median forebrain bundle ([@B2]). Taken together, studies in TAAR1 KO animals support the role of TAAR1 as a regulator of dopaminergic neurotransmission, underlining the role of TAAR1 as a potential novel target for the treatment of neuropsychiatric disorders.
Here we extended the studies of TAs, thyronamines and TAAR1 on dopamine neurotransmission in the dorsal striatum. Special emphasis was put on T~1~AM, which contains the aryethylamine backbone of monoamine neurotransmitters ([@B17]). Remarkably, T~1~AM is a product of the enzymatic deiodination and decarboxylation of T~4~ ([@B40]). We used slices from WT and TAAR1 KO mice and examined the effects of tyramine, β-PEA, and T~1~AM on the phosphorylation state of tyrosine hydroxylase (TH), which regulates DA synthesis ([@B20]), along with TH activity. For further evaluation of TH activity, we measured with high pressure liquid chromatography (HPLC) the levels of [L]{.smallcaps}-DOPA after the administration of a DOPA decarboxylase inhibitor. Using the same slices, we also studied effects of T~1~AM on evoked DA release. We also studied effects on the phosphorylation state of the post-synaptic alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit GluA1, which plays a crucial role in regulating transmission and plasticity at excitatory synapses in striatum. Finally, mass spectrometry imaging was used to detect T~1~AM at baseline and upon systemic administration.
Materials and Methods {#s1}
=====================
Animals
-------
The experiments were approved by the local ethical committee at Karolinska Institute (N351/08) and conducted in accordance with the European Communities Council Directive of 24 November 1986 (86/609/EEC). Adult male WT and TAAR1 KO mice on a C57Bl6 background were used ([@B22]). They were housed in temperature- and humidity-controlled rooms (20°C, 53% humidity) with a 12 h dark/light cycle. They had access to standard lab pellets and water *ad libitum*.
Preparation and Incubation of Dorsal Striatal Slices for Phosphorylation Experiments
------------------------------------------------------------------------------------
Mouse brains were rapidly removed and placed in ice-cold, oxygenated (95% O~2~/5% CO~2~) artificial cerebrospinal fluid (aCSF) containing (in mM): 126 NaCl, 2.5 KCl, 1.2 NaH~2~PO~4~, 1.3 MgCl~2~, 2.4 CaCl~2~, 10 glucose and 26 NaHCO~3~, pH 7.4. Coronal slices (300 μm thick) were prepared using a Leica vibratome (Leica, Wetzlar, Germany). Dorsal striata were dissected from the slices in ice-cold aCSF buffer. Each slice was placed in a polypropylene incubation tube with 2 ml fresh aCSF buffer. The slices were preincubated at 30°C under constant oxygenation (95% O~2~/5% CO~2~) for 60 min with a change of buffer after 30 min. The buffer was then replaced with fresh aCSF and slices were treated with tyramine (1, 10, 100 μM; Sigma-Aldrich, St. Louis, MO, United States), T~1~AM (1, 10, 100 μM; synthesized by Servier, kind gift from Mark J. Millan), β-PEA (100 μM; Sigma-Aldrich), SCH23390 (5 μM; Sigma-Aldrich), EPPTB (10 nM, synthesized by Servier, kind gift from Mark J. Millan), KN-92 (10 μM; Sigma-RBI), H-89 (10 μM; Calbiochem, Gibbstown, NJ, United States) and 8-CPT-2Me-cAMP (10 μM; Tocris Bioscience, Bristol, United Kingdom), alone or in combination. The higher doses of all compounds exceeds by far the IC~50~ or Kd values for their respective target, but it is known that much higher concentrations are needed to exert actions in brain slices when compared to cell culture systems ([@B66]). After the drug treatment, the buffer was removed, the slices rapidly frozen on dry ice and stored at -80°C until assayed.
Immunoblotting
--------------
Immunoblotting was performed as described earlier ([@B71]). Frozen tissue samples were sonicated in 1% SDS, transferred to Eppendorf tubes and boiled for additional 10 min. Small aliquots of the homogenate were retained for protein determination using the bicinchoninic acid protein assay method (Pierce, Rockford, IL, United States). Equal amounts of protein (20 μg) were loaded onto 12% acrylamide gels, and the proteins were separated by SDS-PAGE and transferred to Immobilon^®^-P Polyvinylidene Difluoride membranes (Sigma). Immunoblotting was performed on the membranes using P-Ser^19^-TH (Merck Millipore, Billerica, MA, United States), P-Ser^31^-TH (Millipore), P-Ser^40^-TH (Millipore), P-Ser^845^-GluA1 (UBI), and antibodies, which are not phosphorylation state-specific to estimate total levels of TH (Millipore) and GluA1 (UBI). The antibody binding was detected by incubation with goat anti-mouse or anti-rabbit horseradish peroxidase-linked IgG (1:6000--8000 dilution) and detected using ECL immunoblotting detection reagents (GE Healthcare, Little Chalfont, United Kingdom).
Determination of [L]{.smallcaps}-DOPA in Dorsal Striatal Slices
---------------------------------------------------------------
Dorsal striatal slices were incubated for 5 min with T~1~AM (10 μM) or tyramine (100 μM), and then for 15 min with T~1~AM or tyramine along with the [L]{.smallcaps}-amino acid decarboxylase inhibitor NSD-1015 (100 μM, Sigma-Aldrich). After the removal of the solutions, tissue slices were frozen and sonicated (10,000 *g* for 10 min) in 100 μL perchloric acid (0.1 mM). The pellets were resuspended in 100 μl 1% sodium dodecyl sulfate and the protein content was determined. The level of [L]{.smallcaps}-DOPA in the supernatant was determined using HPLC coupled to an electrochemical detection system with a refrigerated microsampling unit (model CMA/200; CMA Microdialysis, Kista, Sweden). The HPLC apparatus comprised an HPLC pump (model 2150; Pharmacia LKB Biotechnology AB, Uppsala, Sweden) that kept a constant flow of 0.2 mL/min of the mobile phase (0.12 m NaH~2~PO~4~H~2~O; 0.09 m EDTA, 0.05 mm 1-octanesulfonic acid, and 15% methanol, pH 4.2) and a pressure of ∼0.5 bar on a reverse-phase ion pair C-18 column prepacked with Biophase ODS 5 μm particles (BAS, West Lafayette, IN, United States). [L]{.smallcaps}-DOPA was detected with an amperometric detector (model LC-4C; BAS) and a glassy carbon electrode set at 0.75 V. The limit of detection was ∼10 nM.
Amperometry in Dorsal Striatal Slices
-------------------------------------
Sagittal striatal brain slices were prepared and maintained as above. Amperometric detection of DA release was performed as described earlier ([@B100]). Carbon fiber electrodes (10 μm in diameter, World Precision Instruments, Hertfordshire, England) had an active part (100 μm) that was positioned within the dorsal striatum in the brain slice. A constant voltage of +500 mV was applied to the carbon fiber via an Axopatch 200B amplifier (Axon Instruments) and currents were recorded with the same amplifier. A stimulating electrode (patch electrode filled with aCSF) was placed on the slice surface, in the vicinity of the carbon fiber electrode. Stimulation consisted of a single pulse (0.1 ms, 8--14 μA) applied every minute, which evoked a response corresponding to oxidation of DA at the surface of the electrode. When the carbon fiber electrode was held at 0 mV, stimulation of the slice did not produce any current.
Matrix-Assisted Laser Desorption Ionization (MALDI) -- Mass Spectrometry (MS) Imaging
-------------------------------------------------------------------------------------
Adult male WT and TAAR1 KO mice were injected with saline or T~1~AM (20 mg/kg, i.p.) and killed by decapitation 30 or 60 min post-dose. All brains were immediately removed, snap frozen, and stored at -80°C until further analysis. The frozen brain tissues were cut using a cryostat-microtome (Leica CM3050S; Leica Microsystems, Welzlar, Germany) at a thickness of 14 μm, thaw-mounted onto conductive indium tin oxide (ITO) glass slides (Bruker Daltonics), and stored at -80°C. Sections were dried gently under a flow of nitrogen and desiccated at room temperature for 15 min, after which they were imaged optically using a photo scanner (Epson perfection V500). The samples were then coated with derivatization reagents, 2, 4-diphenylpyrylium tetrafluoroborate (DPP-TFB). Stock solution of DPP-TFB (8 mg in 1.2 ml MeOH) was prepared and diluted in 6 mL of 70% methanol containing 3.5 μL of trimethylamine. An automated pneumatic sprayer (TM-Sprayer, HTX Technologies, Carrboro, NC, United States) was used to spray DPP-TFB solution over the tissue sections. The nozzle temperature was set at 80°C and the reagent was sprayed for 30 passes over the tissue sections at a linear velocity of 110 cm/min with a flow rate of about 80 μL/min. Samples were then incubated for 15 min (dried by nitrogen flow every 5 min) in a chamber saturated with vapor from a 50% methanol solution. MALDI-MSI experiment was performed using a MALDI-TOF/TOF (Ultraflextreme, Bruker Daltonics, Bremen, Germany) mass spectrometer with a Smartbeam II 2 kHz laser in positive ion mode. The laser power was optimized at the start of each run and then held constant during the MALDI-MSI experiment.
Data Analysis and Statistics
----------------------------
Autoradiograms from western blotting experiments were digitized using a Dia-Scanner (Epson Perfection 4870 PHOTO). Optical density values were measured using NIH Scion Image for Windows (alfa 4.0.3.2; © 2000--2001 Scion Corporation). Biochemical data were analyzed using one-way ANOVAs followed by Newman--Keuls *post hoc* test. Data from Amperometry were acquired and analyzed with the pClamp 9 or pClamp 10 software (Axon Instruments). Data are expressed as % of the baseline response measured for each slice during the 5--10 min preceding start of perfusion with T~1~AM. Statistical significance of the results was assessed by using Student's *t*-test for paired observations (comparisons with baseline within single groups) or one-way [ANOVA]{.smallcaps} multiple comparison test followed by Newman--Keuls *post hoc* test since the samples from WT and KO mice were loaded in separated gels. The numbers of individual replicates are shown in the graphs, while *p*-values, degrees of freedom and *F* values are detailed in the results part.
Results
=======
Dose Responses of Tyramine and T~1~AM on Phosphorylation of TH and GluA1 in Striatal Slices From TAAR1 Receptor WT and KO Mice
------------------------------------------------------------------------------------------------------------------------------
We first studied the dose responses of the tyramine and T~1~AM on the phosphorylation of TH and GluA1 in striatal slices from TAAR1 receptor WT and KO mice. To study effects of compounds in the both genotypes, their individual baseline was set at 100%. One way ANOVA analysis showed that tyramine caused a significant change of P-Ser^40^-TH in both WT (*F*~\[3,20\]~: 60.28; *p* \< 0.0001) and KO mice (*F*~\[3,20\]~: 9.928, *p* = 0.0003). *Post hoc* test showed that the lower (1 and 10 μM) concentrations of tyramine did not have any significant effect, whereas 100 μM tyramine significantly reduced phosphorylation of P-Ser^40^-TH in both groups of mice, suggesting an effect independent of TAAR1 (**Figure [1A](#F1){ref-type="fig"}**). Meanwhile, there were no effects at the same concentration of tyramine on P-Ser^19^-TH (WT: *F*~\[3,20\]~: 2.913; *p* = 0.06; KO: *F*~\[3,20\]~: 2.718; *p* = 0.07) and P-Ser^31^-TH (WT: *F*~\[3,20\]~: 0.9955; *p* = 0.42; KO: *F*~\[3,20\]~: 0.9157; *p* = 0.45) in neither WT nor KO mice (**Figure [1A](#F1){ref-type="fig"}**). One way ANOVA analysis in T~1~AM treated mice revealed that only in the WT but not in the KO mice, the drug affected significantly the levels of P-Ser^19^ (WT: *F*~\[3,20\]~: 3.641; *p* = 0.0004; KO: *F*~\[3,20\]~: 0.1953, *p* = 0.8983) and P-Ser^40^-TH (WT: *F*~\[3,16\]~: 4.393, *p* = 0.02; KO: *F*~\[3,16\]~: 0.823, *p* = 0.5001). *Post hoc* test showed that 10 μM T~1~AM enhanced P-Ser^19^ and P-Ser^40^-TH in the striatum of WT but not TAAR1 KO mice (**Figure [1B](#F1){ref-type="fig"}**). The effects of T~1~AM were biphasic with a further increase in drug concentrations resulting in less phosphorylation of TH. There was no effect of T~1~AM on P-Ser^31^-TH both in WT (*F*~\[3,20\]~: 0.7911, *p* = 0.5131) and KO mice (*F*~\[3,20\]~: 0.446, *p* = 0.7228) (**Figure [1B](#F1){ref-type="fig"}**). Tyramine altered significantly the phosphorylation of P-Ser^845^-GluA1 in WT (*F*~\[3,20\]~: 9.387; *p* = 0.0004) and KO mice (*F*~\[3,20\]~: 22.89; *p* \< 0.0001) (**Figure [1C](#F1){ref-type="fig"}**). In contrast, T~1~AM had no effect on P-Ser^845^-GluA1 in striatal slices from either TAAR1 WT (*F*~\[3,20\]~: 0.8744; *p* = 0.4709) or KO mice (*F*~\[3,20\]~: 0.4367; *p* = 0.7295) (**Figure [1D](#F1){ref-type="fig"}**). These data suggest that tyramine and T~1~AM act differently on pre- and post-synaptic striatal targets.
{#F1}
To further study the effect of a high dose of endogenous TAs, we incubated striatal slices from WT mice with the tyramine (100 μM), β-PEA (100 μM) alone or with the D~1~ receptor antagonist SCH23390 (5 μM). As shown in **Figure [2](#F2){ref-type="fig"}**, we found that β-PEA, like tyramine, decreased P-Ser^40^-TH (*F*~\[5,18\]~: 68.88; *p* \< 0.0001) while not significantly affecting P-Ser^19^-TH (*F*~\[5,18\]~: 1.442; *p* = 0.2573) or P-Ser^31^-TH (*F*~\[5,18\]~: 1.18; *p* = 0.3574). The effects of tyramine and β-PEA on P-Ser^40^-TH were not affected by SCH23390. On the other hand, tyramine significantly enhanced P-Ser^845^-GluA1 (*F*~\[5,18\]~: 2.455; *p* \< 0.0001), an effect that was reversed to baseline by D~1~ receptor blockade using SCH23390. Likewise, β-PEA tended to increase P-Ser^845^-GluA1, but this effect did not reach significance. The data of β-PEA was further to confirm that classical TAs and T~1~AM act differently.
{#F2}
Effects of Tyramine and T~1~AM on TH Activity Measured by [L]{.smallcaps}-DOPA in Striatal Slices From TAAR1 WT and KO Mice
---------------------------------------------------------------------------------------------------------------------------
There was also a baseline increase of TH activity in TAAR1 KO mice as compared to WT mice ([@B22]). To study effects of compounds in the both genotypes, their individual baseline was set at 100%. One way ANOVA revealed significant difference among the groups in WT mice (*F*~\[2,18\]~: 6.856; *p* = 0.0061) (**Figure [3](#F3){ref-type="fig"}**). *Post hoc* test showed that T~1~AM (10 μM), but not tyramine (100 μM), induced [L]{.smallcaps}-DOPA accumulation in the presence of a DOPA decarboxylase inhibitor in WT mice. No effects were detected in TAAR1 KO mice (*F*~\[2,18\]~: 0.08823; *p* = 0.9159), indicative of a TAAR1-mediated mechanism of action of T~1~AM.
![Effect of tyramine and T~1~AM on TH activity measured by [L]{.smallcaps}-DOPA in striatal slices from WT and TAAR1 KO mice. The activity of TH as measured by [L]{.smallcaps}-DOPA was enhanced by T~1~AM (10 μM) in WT mice whereas no change was detected in TAAR1 KO mice. Tyramine (100 μM) had no effect on [L]{.smallcaps}-DOPA. ^∗∗^*p* \< 0.01; one-way ANOVA followed by Newman--Keuls test for pairwise comparisons. The number of individual replicates is indicated within each column.](fphar-09-00166-g003){#F3}
Effects of T~1~AM Alone or in Combination With H-89, 8-CPT-2Me-cAMP, and KN-92 on Phosphorylation of TH in Dorsal Striatal Slices From WT Mice
----------------------------------------------------------------------------------------------------------------------------------------------
To further study intracellular signaling cascades underlying the T~1~AM-induced phosphorylation of TH, we combined the T~1~AM with KN-92 or H-89, inhibitors of CamKII and protein kinase A (PKA), respectively. The effects of T~1~AM on both P-Ser^19^ (*F*~\[7,75\]~: 3.651; *p* = 0.0019) and P-Ser^40^-TH (*F*~\[7,75\]~: 3.871; *p* = 0.0012) could be significantly inhibited by either KN-92 or H-89 (**Figure [4](#F4){ref-type="fig"}**). Since TAAR1 is a Gs-coupled receptor and generates cAMP, we also examined the effects of 8-CPT-2Me-cAMP, an EPAC (Exchange Protein directly Activated by cAMP) activator, on TH phosphorylation. 8-CPT-2Me-cAMP alone tended to increase TH phosphorylation, but did not interact with T~1~AM (**Figure [4](#F4){ref-type="fig"}**).
{#F4}
Effects of T~1~AM and EPPTB, Alone or in Combination, on Evoked Dopamine Release and Phosphorylation of TH in Dorsal Striatal Slices From WT and TAAR1 KO Mice
--------------------------------------------------------------------------------------------------------------------------------------------------------------
We evaluated the effect of T~1~AM on stimulation-evoked release of DA from DA-containing fibers present in sagittal striatal slices, as shown in **Figure [5](#F5){ref-type="fig"}**. We found that bath application of T~1~AM (10 μM) significantly increased the amplitude of evoked DA release measured with carbon fiber electrodes coupled to amperometry in dorsal striatal brain slices from WT mice, and that this effect was significantly reduced in TAAR1 KO mice. In presence of the TAAR1 antagonist EPPTB (10 nM), the effect of T~1~AM on evoked DA release in WT mice was significantly reduced, but not completely blocked (*F*~\[2,20\]~: 7,252; *p* = 0.0043). EPPTB had no effect on evoked DA release by itself (data not shown). Similar to the results as above, T~1~AM significantly increased P-Ser^40^-TH (*F*~\[3,30\]~: 3.384; *p* = 0.0309). This effect was blocked when T~1~AM was combined with EPPTB, suggesting a mechanism of action mediated via TAAR1.
{#F5}
The Relative Distribution and Abundance of T~1~AM in Sagittal Brain Sections From WT and TAAR1 KO Mice
------------------------------------------------------------------------------------------------------
T~1~AM was derivatized by 2, 4 diphenyl pyranylium and detected by MALDI-MSI. No clear endogenous signal of T~1~AM was found in uninjected sections neither from WT nor TAAR1 KO mice. However, widespread signals corresponding to derivatized T~1~AM was detected in mice intraperitoneally administered with T~1~AM (20 mg/kg). The concentration of the drug appeared higher after 30 min compared to 60 min post-dose (**Figure [6](#F6){ref-type="fig"}**).
{#F6}
Discussion
==========
Our experiments demonstrate that the thyronamine T~1~AM enhance phosphorylation and activity of TH along with evoked DA release in dorsal striatum, while not significantly affecting the phosphorylation of post-synaptic AMPA receptor GluA1 subunits. The effects on TH phosphorylation observed following T~1~AM administration were abolished in TAAR1 KO mice, while the effects on evoked DA release were attenuated in TAAR1 KO mice and following TAAR1 blockade, supporting a role of TAAR1 as a partial mediator of these effects. We can conclude that T~1~AM acts through TAAR1 to enhance the production of dopamine. This *de novo* dopamine creation heightens the synaptic dopamine content and raises extracellular dopamine. In contrast, tyramine and β-PEA reduced TH phosphorylation via a mechanism independent of TAAR1.
The Differential Effects of the TAs Tyramine, β-PEA and Thyronamine, T~1~AM, on Phosphorylation and Activity of TH and on GluA1 Phosphorylation
-----------------------------------------------------------------------------------------------------------------------------------------------
Alterations of DA synthesis are regulated via phosphorylation of TH, the rate-limiting enzyme in the synthesis of catecholamines ([@B20]), and Ser^19^, Ser^31^, and Ser^40^ have been identified as the functionally most important sites of TH phosphorylation ([@B39]). Phosphorylation of Ser^19^ is induced by enhanced intracellular Ca^2+^ concentrations and activation of CaM kinase II, whereas phosphorylation at Ser^31^ is induced by extracellular signal-regulated protein kinases, and phosphorylation of Ser^40^ is catalyzed by PKA ([@B37]). TH phosphorylation at Ser^40^ and Ser^31^ leads to increased TH activity, whereas phosphorylation at Ser^19^ exerts a positive modulatory influence on Ser^40^ phosphorylation ([@B7]; [@B27]). Notably, TAAR1 KO animals exhibit a basal increase in TH activity and increased basal phosphorylation at Ser^19^, Ser^31^, and Ser^40^ in striatal slices compared to WT animals, perhaps due to developmental compensations ([@B22]). An increased TH activity at Ser^19^, Ser^31^, and Ser^40^ was observed in WT mice following administration of 1 to 10 μM T~1~AM, but this effect was attenuated upon increasing or lowering the concentration, indicating a bell-shaped dose--response. The effects of T~1~AM on TH activity were abolished in TAAR1 KO mice, identifying TAAR1 as a mediator of these actions. TAAR1 is coupled with stimulatory G~s~ proteins as well as G protein independent pathways and, upon activation, TAAR1 signals through the cAMP/PKA/CREB, β-arrestin2/Akt/GSK-3β and the protein kinase C (PKC)/Ca++/NFAT pathways ([@B8]; [@B14]; [@B67]; [@B35]). Here we found that the T~1~AM-mediated increases on Ser^19^ and Ser^40^ TH were inhibited by blockade of CamKII and PKA by either KN-92 or H-89, respectively. The protein responsible for the phosphorylation of TH at the site Ser^31^ is MAPK ([@B38]). As a consequence we suppose that the TAAR1's activation by T~1~AM stimulates the activity of MAPK, either through PKA or through an alternative direct pathway like β-arrestin2 ([@B69]). In order to confirm that the observed phosphorylation of TH leads to increased enzymatic activity, we made measurements of DOPA with HPLC in striatal slices. T~1~AM induced a higher level of DOPA accumulation in the presence of a DOPA decarboxylase inhibitor in WT mice. This effect of T~1~AM was abolished in KO counterparts. **Figure [7](#F7){ref-type="fig"}** shows a schematic, and somewhat speculative, drawing of the proposed signaling pathway induced by T~1~AM/TAAR1/PKA activation. TAAR1 is a G~s~ protein-coupled receptor and activation results in increased cAMP via activation of adenylyl cyclase and PKA signaling, which directly phosphorylates TH at Ser^40^. Another pathway could involve activation of inhibitor 1 by PKA, which inhibits protein phosphatase 1 (PP-1). PP-1 inhibits the phosphorylation of CamKII, which activates phosphorylation of Ser^19^-TH. CamKII can also inhibit protein phosphatase 2A which activates phosphorylation of Ser^40^-TH. The regulation of the suggested signal transduction pathways could explain how KN-92 and H-89, inhibitors of CamKII and PKA respectively, both can block the effect of T~1~AM. 8-CPT-2Me-cAMP, the EPAC analog, had no influence on TH phosphorylation. In summary, this higher phosphorylation rate leads to the accumulation of DOPA.
{#F7}
In contrast to T~1~AM, under the present conditions, the endogenous TAs tyramine and β-PEA appeared to elicit mainly non-TAAR1 dependent effects. Tyramine, at a high dose, reduced phosphorylation of Ser^40^-TH, but increased phosphorylation of the post-synaptic AMPA receptor GluA1 subunit in striatal slices from both WT and TAAR1 KO mice. Similarly, another endogenous TA, β-PEA also reduced phosphorylation of Ser^40^-TH and tended to increase phosphorylation of Ser^845^-GluA1 in slices from WT mice. The quantification of GluA1 phosphorylation on Ser^845^, help us to evaluate the status of GluA1 in the post-synaptic membrane of corticostriatal and thalamostriatal synapses, which are the main classes of glutamatergic synapse in the striatum ([@B83]). The corticostriatal and thalamocortical projection neurons innervate the principal population of MSNs, but also various subtypes of interneurons ([@B83]). By measuring the phosphorylation of GluA1 we cannot rule out which types of interneurons and MSNs (D~1~ positive or D~2~ positive) are activated. Nevertheless, it is shown that D~1~ agonists and D~2~ antagonists induce robust increases in GluA1 phosphorylation while D~2~ agonists and D~1~ antagonists have no effect ([@B99]). Considering this fact, we deduce that the effect of tyramine on GluA1 trafficking could be explained by either the post-synaptic regulation of D~1~ or D~2~ receptors. D~1~ receptor blockade by SCH23390 blocked the tyramine-induced phosphorylation of GluA1 subunits, but had no effect on tyramine- or β-PEA-reduced Ser^40^-TH phosphorylation, suggesting that the post-synaptic effects of tyramine are dependent on D~1~-receptor activation. According, GluA1 upregulation may be a consequence of the dopamine's net effect upon D~1~ and D~2~ receptors ([@B99]). In our study, we observed that tyramine attenuated the TH phosphorylation at Ser^40^. However, diminished enzymatic activity of TH does not mean a reduction in dopamine release, whilst a negative feedback mechanism for dopamine control of TH activity has been documented ([@B56]; [@B20]). It is conceivable that tyramine has amphetamine-like effects on the excitability of the post-synaptic membrane and possibly leads to the vesicular leak of dopamine by its interaction with VMAT2 ([@B102]). Moreover, it has been suggested that TAs act like amphetamines and could increase extracellular DA levels by promoting DA release via inducing reversal of the dopamine transporter (DAT) and by displacing DA from vesicular stores ([@B89]; [@B43]; [@B64]). β-PEA, which is structurally related to amphetamine, has been proposed to act as an endogenous amphetamine ([@B42]), and has previously been shown to increase extracellular levels of DA in striatum and NAc via a DAT-dependent mechanism ([@B84]; [@B65]). Notably, Xie and Miller found that TAs, including tyramine and β-PEA, do not directly activate monoamine autoreceptors ([@B96]). However, they have been proposed to indirectly activate dopamine autoreceptors by enhancing the efflux of dopamine ([@B32]). One possible explanation for tyramine's TAAR1 independent effect on GluA1 phosphorylation, may be that this TA acts through MSN-localized TAAR1 to affect the availability of GluN1 and through VMAT2 to alter the surface density of GluA1. Indeed, several studies have supposed that tyramine can affect glutamate receptor membrane availability through MSN-localized TAAR1 ([@B2]; [@B28]; [@B88]). To conclude, our data support the notion that T~1~AM can modulate DA synthesis via a mechanism of action that involves presynaptic TAAR1. Moreover, we suppose a direct effect of tyramine on dopamine release that could lead to the observed decline in TH phosphorylation due to secondary activation of indirect D~2~ autoreceptors ([@B56]).
The Effect of T~1~AM on Evoked DA Release in the Dorsal Striatum Using Amperometry
----------------------------------------------------------------------------------
Our experiments suggest that striatal dopamine release can be enhanced by T~1~AM-mediated TAAR1 activation. However, most previous slice experiments addressing the modulatory influence of TAAR1 on the dopaminergic system have been performed in the VTA ([@B55]; [@B72]), which may be a source of discrepancy between our study and previous findings. Previous slice experiments in the VTA of TAAR1 KO mice revealed enhanced spontaneous firing rates of dopaminergic neurons in TAAR1 KO mice compared to WT mice, suggesting that TAAR1 exerts an attenuating effect on dopaminergic neuron activity ([@B55]). This has been supported by slice experiments in mouse VTA using the specific ligands RO5166017 and EPPTB to stimulate and block TAAR1, respectively ([@B10]; [@B73]). However, the increased DA neuron firing rate observed in TAAR1 KO mice did not lead to enhanced basal levels of extracellular striatal DA compared to WT mice as detected by microdialysis ([@B55]). Indeed, mice overexpressing TAAR1, like TAAR1 KO mice, also exhibit an enhanced spontaneous firing activity of monoaminergic neurons of the VTA, DRN, and locus coeruleus ([@B72]). Moreover, it is likely that the functional outcome of TAAR1 activation differs between specific classes of ligand and distinct brain regions depending on the characteristics of the dopaminergic innervation and basal tone. Although midbrain DA neurons are considered to be relatively homogenous, emerging data support a high level of diversity among VTA and SNc neurons as regards electrophysiological properties, synaptic connectivity, protein expression profiles, and behavioral functions ([@B75]). The expression levels of two TAAR1 related proteins, D~2~ receptor and GIRK2, are implicated in the differences between the two subpopulations ([@B10]; [@B74]; [@B6]; [@B12]). In agreement with this, it is reported that TAAR1 has a differential role in dopamine release between VTA and SNc projection sites in striatum ([@B53]). In contrast, [@B22] showed that TAAR1 decreases the amplitude of Methylenedioxymethamphetamine (MDMA) induced dopamine release both in ventral and dorsal striatum. In the same study it was observed that the TAAR1 agonist, *o*-phenyl-3-iodotyramine (*o*-PIT) blunted the para-chloroamphetamine (PCA) induced dopamine release in both structures ([@B22]). Accordingly, TAAR1 may exert a complex pattern of effects on dopaminergic terminals in ventral as compared to dorsal compartments of the striatum. Furthermore, both the VTA and SNc can be further distinguished regarding the expression of calbindin D28k (CB), with the highest density of CB positive neurons located in VTA ([@B90]). CB positive neurons have the tendency to send projections in CB poor islands in striatum (striosomes), while CB negative cells mainly innervate CB rich regions of the striatum (striatal matrix) ([@B13]; [@B81]). It has been reported that the evoked striatal DA release differs between these two compartments but also that the dopamine release ratio of striosome over matrix is higher in the ventral than dorsal striatum ([@B78]). Consequently, TAAR1 could have diverse effects not only among VTA and SNc neurons but also between CB positive and negative subgroups.
Our findings with T~1~AM, may also be explained by differences in the methodologies and protocols employed to evoke and measure dopamine release. For example, the use of strong stimulation intensities might evoke maximal release. Conversely, local, low intensity stimulation, as used in the present study, allows for observation of both inhibition and potentiation of dopamine release. In addition, recent studies have demonstrated that dopamine release in brain slices can be evoked by direct stimulation of dopaminergic axons and indirectly by stimulation of cholinergic interneurons in the striatum ([@B93]; [@B101]). It has not yet been established whether cholinergic neurons express TAAR1 but the contrasting effects of RO5166017 and T~1~AM might result from differences in the involvement of cholinergic control of dopamine release between different experimental paradigms. Taken together, these studies raise the question of a possible differential control of dopamine release by TAAR1 receptors in cholinergic interneurons and in dopamine axon terminals.
In this study, we investigated the action of T~1~AM at TAAR1 on dopaminergic terminals as compared to those of TAs. However, T~1~AM is also known to be an agonist of TAAR5 ([@B25]). Moreover, the β-phenylethylamine-like structure affords T~1~AM the ability to bind with various members of GPCR superfamily and ion channels ([@B17]; [@B44]). It is indeed claimed that T~1~AM interacts with α2a adrenergic receptors, β2-adrenergic receptors and muscarinic receptors ([@B46]; [@B23],[@B24]; [@B51], [@B50]). Notably, outside the CNS, T~1~AM has been found to differentially regulate insulin secretion through actions at TAAR1 and α2a adrenergic receptor ([@B17]; [@B52]). Hence, despite blockade of the actions of T~1~AM in KO mice and by pharmacological antagonist, the possibility that it exerts actions via other mechanisms should not be excluded.
In this study we incubated the slices in a T~1~AM containing buffer. It is important to access the roles of T~1~AM in the intact brain. We show here that T~1~AM can be detected by MALDI-MSI in mouse brain slices 30 and 60 min after systemic administration. Since T~1~AM was detected in many brain areas, we can conclude that T~1~AM can penetrate the blood brain barrier. This finding is in accordance with previous studies showing effects on glucose metabolism by intraperitoneally administered T~1~AM ([@B47]). Using MALDI-MSI, no clear endogenous levels of T~1~AM could be detected. However, it will be interesting to study T~1~AM levels in pathological states, particularly in hyperthyroid conditions. In addition to a circulating source, direct enzymatic transformation of T~4~ to T~1~AM may occur in neurons. The responsible enzyme for this reaction is ornithine decarboxylase ([@B40]), which is expressed by neuronal and astroglial cell types of the CNS ([@B3]). Apart from T~1~AM itself, its metabolite 3-iodothyroacetic acid (TA1) is implicated in the modulation of histaminergic neurotransmission and might likewise interact with dopaminergic pathways: this remains to be clarified ([@B49]).
Conclusion
==========
This study demonstrates that TAAR1 mediates the effects of T~1~AM on dorsal striatal TH phosphorylation, activity and evoked dopamine release. No comparable alterations were found after application of tyramine and β-PEA. This simultaneous augmentation in TH phosphorylation and striatal dopamine release after the administration of T~1~AM indicates that this thyronamine favors dopamine synthesis and subsequent secretion through TAAR1. Conversely, TAs act in a TAAR1 independent manner to influence dopamine secretion resulting in feedback inhibition of TH. This study further indicates that the modulatory properties of TAAR1 may differ depending on the identity of the ligand in question, the extracellular milieu, basal levels of monoamines, neuronal circuitry, and the cellular localization of TAAR1, which are mutually regulated by interactions with D~2~ receptors and DAT, and by the available signaling transduction systems. Further elucidation of the complex pattern of influence of TAAR1 upon monoaminergic and other pathways controlling mood, motor function and cognition may lead to the elaboration of urgently-need, novel strategies for improving the treatment of depression, schizophrenia, Parkinson's disease, and other neuropsychiatric disorders ([@B60]; [@B45]; [@B5]; [@B19]).
Author Contributions
====================
Participated in research design: XZ, KC, and PS. Collected the samples and conducted the experiments: XZ, MS, MP, AN, and TY. Performed the data analysis and discussed the data: XZ, IM, AA, TY, JK, PEA, MJM, KC, and PS. Contributed to the writing of the manuscript and to revising it critically for scientific discussions: XZ, IM, AA, MJM, KC, and PS. All authors approved the final version to be published.
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.
**Funding.** This study was supported by Konung Gustaf V:s och Drottning Victorias Frimurarestiftelse and the Swedish Research Council to PS. PEA was supported by the Swedish Research Council (Medicine and Health \#2013-3105, Natural and Engineering Science \#2014-6215), the Swedish Brain Foundation, the Swedish Foundation for Strategic Research \#RIF14-0078 and Science for Life Laboratory (SciLifeLab). MP got a scholarship from CAPES -- Coordination for the Improvement of Higher Education Personnel.
[^1]: Edited by: *Stefano Espinoza, Fondazione Istituto Italiano di Technologia, Italy*
[^2]: Reviewed by: *Lucia Caffino, Università degli Studi di Milano, Italy; Grazia Chiellini, University of Pisa, Italy*
[^3]: This article was submitted to Neuropharmacology, a section of the journal Frontiers in Pharmacology
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Protein-protein interactions participate in myriad processes of the cell such as replication, transcription, translation, signal transduction, immune response, metabolism, membrane-associated processes and development (e.g., \[[@B1]-[@B4]\]). Protein-protein interactions offer an excellent way of combining its limited working parts, the proteins, to achieve large functional diversity using a limited genetic repertoire \[[@B5]\]. Abnormal interactions between proteins within the cell or from pathogens cause many human diseases \[[@B6]\]. Protein binding can also elicit an allosteric response. Allostery is an integral and pervasive mechanism employed by nature to modulate cellular processes \[[@B7]-[@B11]\]. It serves as a key mechanism for obtaining fine-tuned regulation in several cellular processes -- from metabolic pathways, signalling systems \[[@B12]\] to gene regulation \[[@B13]\]. Functional modulation is achieved either by enhancing (positive co-operativity) or decreasing (negative co-operativity) levels of function. The effect at target site can be varied, e.g., activation of catalysis, regulation of ligand-binding, control of complex formation\[[@B9]\].
Given their importance, several high-throughput interaction assays \[[@B14],[@B15]\], such as yeast two-hybrid and tandem affinity purification, have been developed to supplement the dataset of protein-protein interactions from low-throughput methods \[[@B16],[@B17]\]. However, such large-scale experimental methods suffer from high false-positive rates \[[@B18]\]. The gold standard for protein-protein interactions is usually a dataset of complexes of interacting proteins solved using X-ray crystallography \[[@B19]-[@B21]\]. Although it is a much smaller and incomplete dataset in comparison to high-throughput protein-protein interaction datasets, it is reliable and enables mapping of interaction regions and structural changes which accompany interactions. Several derived databases provide protein-protein interaction datasets in various easy to --study and --use formats. SCOPPI \[[@B22]\], iPfam \[[@B23]\], SNAPPI-DB \[[@B24]\], 3D Complex \[[@B25]\], InterEvol \[[@B26]\] and ProtCID \[[@B27]\] are some of the available 3D structural databases of protein-protein complexes.
Protein-protein interactions can be classified into different kinds \[[@B28]\]: homo-oligomers and hetero-oligomers; obligate and non-obligate complexes; permanent and transient complexes. Non-obligate complexes form an important class since they serve as key regulators in maintaining and regulating cellular homeostasis \[[@B29]-[@B31]\]. They are also valuable from the viewpoint of structural biology since both the unbound and bound forms can be crystallized owing to their stability. Several such structures have been solved by various groups and deposited in the Protein Data Bank (PDB) \[[@B32]\]. An invaluable non-redundant dataset of structures of both the interacting partners solved in unbound and bound form has been collated, curated and updated by Weng and colleagues \[[@B33],[@B34]\]. The ComSin database provides a unique collection of structures of proteins solved in unbound and bound form, targeted towards disorder--order transitions \[[@B35]\].
Earlier studies of structures of protein-protein complexes using both the unbound and bound form of proteins reveal that proteins undergo changes in their structure upon binding. Betts and Sternberg \[[@B36]\] were the first to compare the bound and unbound forms using a dataset of 39 complexes. Martin et al. \[[@B37],[@B38]\] analyzed a dataset of 83 complexes in terms of local structural variations. The alterations in structure as a result of protein-protein interactions manifest either as a rigid-body shift of a segment or as a conformational change from one secondary structural form to another \[[@B39]\]. The extent of conformational change observed at the interface upon binding prompted several studies to understand and predict these changes \[[@B40]-[@B42]\]. Such studies aim to improve protein-protein docking methods \[[@B43]\] and help in the accurate docking of protein-protein interactions, which can be used to understand the mechanism of functioning of the complex or design inhibitors.
In this work, we have used a curated and non-redundant dataset of 76 protein-protein complexes, solved using X-ray crystallography in high resolution in both unbound and bound form, to address questions about the nature, extent and location of structural changes upon binding. We noticed that, in addition to changes in the interface, possibly allosteric changes causing structural alteration occur in about half of the complexes, indicating a much higher prevalence of this phenomenon caused due to protein binding than appreciated before.
Results
=======
Proteins bound to other proteins undergo larger structural changes than unliganded proteins
-------------------------------------------------------------------------------------------
Structural change observed in different forms of a protein could be due to experimental artifacts \[[@B44]\], intrinsic flexibility \[[@B45]\] or due to a biologically important external perturbation \[[@B46]\], such as ligand binding or post-translational modification. To differentiate structural changes potentially related to protein-protein interactions from those which are artefacts, we compared variations occurring in the dataset of protein-protein complexes with two control datasets (see Additional file [1](#S1){ref-type="supplementary-material"}: Table S1). The first control set (named Control -- Rigid) consists of 50 structures, solved at a resolution ≤2.5 Å, of two fairly rigid and extensively studied proteins: bovine ribonuclease A and sperm whale myoglobin, and provides an indicator of co-ordinate uncertainties. The second control set (named Control -- Monomer) consists of a non-redundant set of 95 clusters of structures of monomers, also solved at a resolution ≤2.5 Å, which serve as a heterogeneous set since this dataset contains both rigid and flexible proteins, thus serving as a control set for understanding intrinsic flexibility. The main dataset of our study named PPC (Protein-protein complexes) is an extensively curated dataset of non-obligatory proteins with their 3-D structures solved in both unbound and bound forms (Additional file [2](#S2){ref-type="supplementary-material"}: Table S2). It consists of 76 non-obligatory complexes representing members of diverse functions (25 enzyme-inhibitor, 11 antigen-antibody and 40 'other' complexes, which largely comprises of signalling proteins). The number of proteins involved in the 76 complexes represent the major SCOP (Structural Classification of Proteins) \[[@B47]\] classes (all α - 32, all β - 84, α/β - 57, α + β - 37). Since the dataset has been pruned to exclude cases with a large percentage of missing residues at the interface, disordered proteins are under-represented. The complexes predominantly involve two-chain interactions and some interactions involving three chains, in which two chains are considered as a single entity (for example, in the case of light and heavy chains of antibody and in the case of G~β~-G~γ~ subunits in heterotrimeric G-proteins). The proteins constitute a mixture of single-domain and multi-domain members. Although some of the structures in the PPC dataset are solved at a resolution poorer than 2.5 Å, the highest resolution of the Control-Rigid and Control-Monomer datasets, the magnitude of structural changes captured across the datasets can be compared since 50/76 complexes of the PPC dataset were solved with a resolution ≤2.5 Å. The conclusions of comparison of various parameters capturing structural change in the different datasets, discussed below, remained unaltered when using either the '50' or '76' set of complexes (data not shown). The conclusions described below are for the entire dataset of 76 complexes.
Three parameters were used to analyze structural change occurring in the different types of residues (see Methods) in a protein: root mean square deviation (RMSD), %PB change (PBc), PB substitution score (PBSSc). Although RMSD captures the magnitude of structural change, it does not distinguish the type of structural change -- i.e. rigid body movement (or) conformational change. The use of Protein Blocks (PBs) enables this distinction since small yet significant changes in local conformation of a protein can be captured using PBs. Protein Blocks consists of 16 standard conformational states of pentapeptides \[[@B48],[@B49]\]. PBs can be used to represent precisely backbone conformation of all the protein structures known so far. This efficient design has been employed in several applications, including prediction of long fragments and short loops, and in identifying proteins with similar structures \[[@B49],[@B50]\]. A PB change between the unbound and bound forms for the equivalent residue indicates a conformational change -- either subtle or drastic. The % PBs altered between two structures serves as a metric for capturing the extent of structural change (PBc). A substitution matrix derived earlier \[[@B50]\] was used to calculate the magnitude of structural dissimilarity between two structures in terms of their PB changes (namely PBSSc). A lower PBSSc indicates unfavourable changes (i.e. drastic conformational change -- for example, a helix to a strand) whereas a higher PBSSc indicates milder conformational changes (for example, change between a curved helix and a linear helix). Analysis of the three parameters revealed that all types of residues (buried, surface, interacting) undergo higher structural change upon binding to another protein than in the unliganded form (Figure [1](#F1){ref-type="fig"}). These values are calculated at per-protein level for the different classes of residues. RMSD (Figure [1](#F1){ref-type="fig"}A) and PBc (Figure [1](#F1){ref-type="fig"}B) clearly showed higher structural variation of protein-bound forms in comparison to the unbound forms whereas PBSSc (Figure [1](#F1){ref-type="fig"}C) showed a marginal trend. This is because the PB changes could be of two kinds: favourable (high PBSSc) and unfavorable (low PBSSc) and both are represented in the graph. As expected, buried residues showed the least deviation of all the classes and interacting residues the highest change. Buried residues are mostly invariant, as seen from the box plot depicting the distribution of PBc (Figure [1](#F1){ref-type="fig"}B), where \~50% of the values are zero for the control datasets. Surprisingly, \~90% of buried residues of protein-protein complex structures show at least a single conformational change, as characterized by change in PB (Figure [1](#F1){ref-type="fig"}B). However, the observed changes are mostly minor. In the rare cases when it is a large change, the residue is seen to have slight exposure to solvent.
{#F1}
In order to distinguish structural variations caused due to protein binding from those occurring due to crystallographic artifacts, the upper bound values corresponding to the Control-Rigid dataset were used as reference for the three parameters (see Additional file [3](#S3){ref-type="supplementary-material"}: Figure S1).
It is observed that the main protein-protein complex dataset comprises of complexes with varying range of interface area and size of proteins (see Additional file [2](#S2){ref-type="supplementary-material"}: Table S2). Therefore, dependence of the parameters capturing structural change for interface area and size of the protein (represented as length of the protein) were analyzed. The analysis indicates that there is slight dependence of RMSD, PBc and PBSSc for interface area buried by the complex whereas the parameters have negligible dependence on the lengths of proteins (see Additional file [4](#S4){ref-type="supplementary-material"}: Figure S2).
Independently, we also captured structural change using all-atom RMSD which includes consideration of sidechain atoms. Although there is expected variation in all-atom RMSDs for interacting residues (see Additional file [5a](#S5){ref-type="supplementary-material"}: Figure S3), there is no profound variation when the Cα RMSDs are compared with corresponding all-atom RMSD values (see Additional file [5b](#S5){ref-type="supplementary-material"}: Figure S3). Therefore, the present analysis is confined to Cα RMSD based comparison in this study.
Pre-made interfaces predominantly bind to structurally-altered interfaces
-------------------------------------------------------------------------
The extent of structural change at both the interfaces of various complexes has been assessed. Proteins are classified into three categories based on the extent of structural change at the interfaces of various complexes: pre-made, induced-fit, and other. Interfaces exhibiting Cα RMSD of \<0.5 Å (which is the maximum deviation between any two proteins of the Control-Rigid dataset) are considered pre-made, while those with Cα RMSD of \>1.5 Å are considered induced-fit. The interfaces showing structural changes between these two values are classified in the 'other' category. We identify 33 pre-made interfaces, fitting the lock-and-key hypothesis proposed to explain protein-ligand binding \[[@B51]\]. Such a large number is surprising since the proteins would always be primed for interaction. Nature's regulatory control of the primed pre-made interfaces appears to be achieved via its partner interface. It appears that although one of the interfaces is pre-made, the other interface undergoes substantial changes (Figure [2](#F2){ref-type="fig"}A, blue coloured points) in the final stable bound form (sometimes \>1.5 Å Cα RMSD, the cut-off for identifying 'induced-fit' interfaces). 9 protein-protein complexes, 5 from 'Other' category and 4 from Enzyme-Inhibitor category, show this behaviour. 17 of the 33 pre-made interfaces had almost no PB changes, implying near complete absence of conformational changes. However, the interaction seems to be modulated by the structural changes occurring in the partner interface. In 14/17 of these cases, the partner's PBc is \>15%; in three cases it is \>40% (Average PBc is 26 ± 14). Only five of the 33 pairs seem to be pre-made in both interfaces. However, inspection of PBc for these 'pre-made interfaces' revealed that there are substantial conformational changes of smaller magnitude captured using PBs which use atomic positions of N, C and O atoms apart from Cα as opposed to Cα-based RMSD (see Figure [2](#F2){ref-type="fig"}B). For instance, the complex of cytochrome C peroxidise and iso-1-cytochrome C forms a pre-made interface (see Additional file [6A](#S6){ref-type="supplementary-material"}: Figure S4), with Cα RMSD of 0.37 Å and 0.31 Å for the interacting proteins. However, the conformations of side chain positions of interface residues changes drastically in 3/18 interface residues. This example supports the hypothesis that almost all interacting partners undergo changes upon binding, even if one of the interfaces is pre-made, and PBs help in identifying subtle changes than classical RMSD measures. In essence, there are no 'completely pre-made' interfaces.
{#F2}
Usually, interfaces with an average Cα RMSD of ≥1.5 Å showed substantial changes at interface, exemplifying the concept of induced-fit hypothesis \[[@B52]\] for formation of protein-protein complexes (see Additional file [6B](#S6){ref-type="supplementary-material"}: Figure S4). 35 interfaces with average Cα RMSD of ≥1.5 Å are found. Of these predominantly altered interfaces, 10 are partners of pre-made interfaces, 4 are partners of like-wise induced-fit interfaces and the rest have values in between (see Additional file [6C,D](#S6){ref-type="supplementary-material"}: Figure S4).
Comparison of the structural change in terms of Cα RMSD and normalized PB substitution score can help in distinguishing cases of rigid body movements from conformational changes. Induced-fit interface regions with 0% PB change at interface can be considered to have rigid-body movements (see Additional file [7A](#S7){ref-type="supplementary-material"}: Figure S5). However, since PBs are very sensitive to backbone torsion angle changes, two very similar PBs will also be considered as PB changes. Therefore, normalized PB substitution score is a more pertinent metric to grade the local conformational change (see Additional file [7B](#S5){ref-type="supplementary-material"}: Figure S5).
Large structural changes could result for different reasons such as to avoid steric clashes and/or optimize binding. In some cases, global changes in the molecule (both interface and non-interacting surface RMSDs are ≥1.5 Å) are observed. These complexes either move out (Figures [3](#F3){ref-type="fig"}A, Additional file [8E](#S8){ref-type="supplementary-material"}: Figure S6) (or) move in (Additional file [8B](#S8){ref-type="supplementary-material"}: Figure S6) to relieve steric clashes/optimize binding, respectively. In most cases, changes were localized at the interface, comprising of rigid-body movements (Additional file [8C](#S8){ref-type="supplementary-material"}: Figure S6) or conformational changes (Additional file [8D](#S8){ref-type="supplementary-material"}: Figure S6) or conformational changes with movement (Additional file [8A](#S8){ref-type="supplementary-material"}: Figure S6), to mainly optimize binding (Figure [3](#F3){ref-type="fig"}B) or relieve steric clashes (Figure [3](#F3){ref-type="fig"}C) or both (Figure [3](#F3){ref-type="fig"}D). Local rearrangements at the interface are identified based on the normalization-based metric (see Methods section). This criterion allows us to identify interfaces with proportionately larger localized changes at the interface although the magnitude is smaller (≤1.5 Å) (Additional file [6D](#S6){ref-type="supplementary-material"}: Figure S4). In cases where the change is larger (≥2 Å), rearrangement seems to be mainly targeted at avoiding steric clashes. In cases where the change is moderate (1.5 Å \~ 2 Å), the rearrangements appear to be mostly for proper optimization of interface.
![**Structural changes observed in interfaces.** Protein undergoing change is shown as cartoon, with unbound form in light cyan and the bound form in blue, and its partner as a ribbon, with unbound form in light orange and bound form in magenta. Direction of movement is indicated as black arrow. **A**) Large (\~10 Å Cα RMSD) *moving out* to avoid steric clash (alpha actin & BNI1 protein; 1Y64). **B**). Movement to optimise interaction with partner (GTP binding protein & Rho GTPase activating protein; 1GRN). **C**). Conformational change accompanied by movement mainly to avoid steric clashes with the partner (Glycoprotein Ib alpha & von Willebrand factor; 1 M10). In B) and **C**), the region of interest is colored green and red in the unbound and bound forms, respectively. **D**). An interface (actin & deoxyribonuclease I; 1ATN) where certain region moves away to avoid steric clash (colored in red), some region undergoes conformational change with movement to optimise an interaction (depicted in green for the unbound form and brown for the bound form) and another region undergoes rigid body movement to optimise its interaction (colored in lemon yellow in the unbound form and orange in the bound form). All the figures containing protein structures were generated using PyMOL \[[@B77]\].](1472-6807-12-6-3){#F3}
Non-interacting regions away from the interface undergo substantial structural changes on binding
-------------------------------------------------------------------------------------------------
In general, interacting residues undergo larger structural change than non-interacting surface residues (≥10% residue surface accessibility (RSA)). Comparison of the three parameters quantifying structural changes studied for individual proteins showed that this trend holds true even in these cases (Figure [4](#F4){ref-type="fig"}, Additional file [9](#S9){ref-type="supplementary-material"}: Figure S7). Indeed, interacting regions need to undergo rearrangement to form an optimal fit. The general trend of comparatively larger changes at interface regions was seen for RMSD and PBSSc (see Additional file [9](#S9){ref-type="supplementary-material"}: Figure S7). However, the parameter PBc provided a new insight, highlighting cases with almost no conformational change at the interface but with considerable change in the rest of the surface (Figure [4](#F4){ref-type="fig"}, green ellipse). This emphasizes that there exist complexes in which non-interacting regions undergo structural variation upon binding even though the interface remains largely unchanged. 6% of the complexes exhibited 10%-25% PBc and one case showed 50% PBc in the non-interacting surface region (Figure [4](#F4){ref-type="fig"}).
{#F4}
Although interacting regions undergo large structural changes in comparison to the rest of the surface, about one-half of the cases in the PPC dataset reveal large changes away from the interface (Figure [4](#F4){ref-type="fig"}, Tables [1](#T1){ref-type="table"} &[2](#T2){ref-type="table"}). PB changes in non-interfacial regions can be divided into two cases. *(i)* Change in non-interacting regions even when there are almost no changes in interacting regions (*n* = 22) (Table [1](#T1){ref-type="table"}). (*ii*) Change in non-interacting regions accompanying changes in interacting regions (*n* = 12) (Table [2](#T2){ref-type="table"}). The two categories combine to provide a dataset of 34/76 complexes exhibiting substantial structural change in non-interfacial surface regions. Interfaces represented in the first case can be considered as pre-made since no PB change is observed on complexation. As expected, the partner protein for these interfaces exhibited much larger change at the interface.
######
Features of proteins with substantial structural change in non-interacting regions and no/moderate change at interface
**PDB code** **Cα RMSD** **Normalized PB substitution score**
------------------------------------- -------------- ------------- -------------------------------------- ----------- ----------------- ------ ----------- -----------------
Protein Bound Unbound IR\* NIR_PBc\* Difference IR\* NIR_PBc\* Difference
(NIR_PBc -- IR) (NIR_PBc -- IR)
PIP3 kinase (O) 1HE8_r 1E8Z 0.49 8.45 7.96 1.25 -1 -2.25
HISF protein (O) 1GPW_r 1THF 0.68 8.10 7.42 1.68 -1.31 -2.99
UCH-L3 (O) 1XD3_r 1UCH 1.30 5.11 3.80 1.35 -1.05 -2.4
Son of Sevenless (O) 1BKD_r 2II0 1.33 4.91 3.57 1.39 -0.82 -2.21
TGF-beta (O) 1KTZ_r 1TGK 0.66 4.41 3.75 2.1 -0.95 -3.05
HPr kinase C-ter domain (E) 1KKL_r 1JB1 1.76 4.00 2.24 1.95 -0.98 -2.93
Cystatin (E) 1YVB_l 1CEW 0.63 3.85 3.21 2.43 -1.24 -3.67
DH/PH domain of TRIO (O) 2NZ8_r 1NTY 1.33 3.77 2.43 1.25 -0.78 -2.03
Actin (O) 2BTF_r 1IJJ 1.08 3.77 2.68 1.76 -0.81 -2.57
Alpha-1-antitrypsin (E) 1OPH_r 1Q1P 0.92 3.50 2.58 1.35 -1.39 -2.74
TGFbeta receptor (O) 1B6C_l 1IAS 1.17 3.14 1.96 1.28 -1.04 -2.32
Vitamin D binding protein (O) 1KXP_l 1KW2 1.53 3.11 1.57 1.66 -0.77 -2.43
TolB (O) 2HQS_r 1CRZ 1.37 3.04 1.67 2 -0.81 -2.81
RCC1 (O) 1I2M_l 1A12 0.40 3.00 2.60 2.43 -0.99 -3.42
Sporulation response factor B (O) 1F51_r 1IXM 0.78 2.70 1.92 1.1 -0.97 -2.07
Ran GTPase (O) 1A2K_l 1QG4 0.33 2.57 2.24 2.56 -1.03 -3.59
HEW lysozyme (A) 1BVK_l 3LZT 0.31 2.57 2.25 2.86 -1.06 -3.92
Transferrin receptor ectodomain (O) 1DE4_l 1CX8 0.94 2.51 1.57 1.04 -1.16 -2.2
Anthrax toxin receptor (O) 1T6B_l 1SHU 0.27 2.32 2.05 2.15 -0.75 -2.9
Xylanase inhibitor (E) 2B42_r 1T6E 0.33 2.25 1.91 2.06 -0.95 -3.01
Fab (A) 1E6J_r 1E6O 0.75 1.98 1.22 2.48 -0.73 -3.21
Complement C3 (O) 1GHQ_r 1C3D 0.21 1.58 1.36 1.92 -0.69 -2.61
\*The abbreviations used are: IR - Interface regions, NIR_PBc -- Non-interacting regions with PB change, l -- ligand (smaller of the two proteins in the complex), r -- receptor (bigger of the two proteins in the complex), E -- enzyme-inhibitor complex, A -- Antigen-antibody complex, O -- Other complexes.
The proteins are listed in decreasing order of average Cα RMSD of NIR_PBc values. The average Cα RMSD of IR and NIR_PBc is 0.89 and 3.75, respectively. The normalized PB substitution score for IR and NIR_PBc values is 1.78 and -0.98, respectively.
######
Proteins with substantial structural change in non-interacting regions and interfacial regions
**PDB code** **Cα RMSD** **Differences from global RMSD** **Normalized PB substitution score**
-------------------------------- -------------- --------- ------------- ---------------------------------- -------------------------------------- --------------------- -------------------------- ------- -----------
Protein Bound Unbound IR\* NIR_PBc\* Global RMSD IR\* -- Global RMSD NIR_PBc\* -- Global RMSD IR\* NIR_PBc\*
Arf1 GTPase (O) 1R8S_r 1HUR 5.19 5.42 3.02 2.17 2.40 0.79 -0.67
Ras GTPase (O) 1BKD_l 1CTQ 4.39 4.43 2.21 2.18 2.22 0.63 -1.82
CDK2 kinase (E) 1FQ1_l 1B39 4.39 4.26 2.04 2.35 2.22 1.31 -1
FC fragment of human IgG 1 (A) 1E4K_r 2DTQ 3.17 3.77 2.18 0.99 1.59 0.95 -0.99
Ran GTPase (O) 1I2M_r 1QG4 3.27 3.51 1.91 1.36 1.60 0.58 -0.8
Cystein protease (E) 1PXV_r 1X9Y 3.99 3.47 1.48 2.51 1.99 1.2 -1.04
Rab21 GTPase (O) 2OT3_l 1YZU 4.65 3.30 1.71 2.94 1.59 0.1 -0.6
CDC42 GTPase (O) 1GRN_r 1A4R 2.66 2.97 1 1.66 1.97 1.31 -0.96
Rac GTPase (O) 2NZ8_l 1MH1 3.87 2.68 1.17 2.70 1.51 -0.03 -1.18
Actin (O) 1ATN_r 1IJJ 6.09 2.58 1.54 4.55 1.04 0.97 -0.59
Rac GTPase (O) 1I4D_l 1MH1 2.31 2.52 0.81 1.50 1.71 1.15 -0.65
Glycoprotein IB-alpha (E) 1M10_l 1MOZ 3.95 2.12 0.89 3.06 1.23 0.91 -0.95
\*The abbreviations used are: *IR* - Interface regions, *NIR_PBc* -- Non-interacting regions with PB change, *l* -- ligand (smaller of the two proteins in the complex), *r* -- receptor (bigger of the two proteins in the complex), *E* -- enzyme-inhibitor complex, *A* -- Antigen-antibody complex, *O* -- Other complexes.
The proteins are listed in decreasing order of average Cα RMSD of NIR_PBc values. The average Cα RMSD of IR and NIR_PBc is 3.99 and 3.42, respectively. The average difference in RMSD values for IR and NIR_PBc from global RMSD are 2.33 and 1.75, respectively. The normalized PB substitution score for IR and NIR_PBc values is 1.737 and −0.95, respectively.
Changes occurring in the non-interfacial regions are classified as near the interface region or away from the interface. All non-interfacial residues in a protein which are within a distance of ≤6 Å Cα distance from any of the interacting residues were considered as 'residues nearby interface', since they occur in the vicinity of the interfacial residues and are important for the formation of the structural scaffold \[[@B53]\]. Figure [5](#F5){ref-type="fig"} shows that in most of the proteins, the residues nearby interface do not undergo much change (Mean -- 15.19%, Median -- 12.31%); the highest peak is at 10%, which means that most of the changes occurred away from the interface. This fact was also confirmed by visual inspection of the structure of the protein-protein complexes.
{#F5}
Conformational changes occurring away from the interface are potentially allosteric: Literature-based, structure-based and normal mode analysis
-----------------------------------------------------------------------------------------------------------------------------------------------
To ascertain any known or potential biological relevance for these changes, all the 'non-interacting regions with PB change' in the identified proteins were analyzed using following parameters: (i) Crystallographic temperature factor (B-factor). Regions with low flexibility and different conformations are likely to have adopted the particular conformation. Studies report that interacting sites have lower B-factors than rest of the protein surface on average even in the unbound form \[[@B54]\]. (ii) Known functional roles of residues: SITE records listed in PDB files and Catalytic Site Atlas (CSA) \[[@B55]\] were consulted to identify if any of the known functionally important residues for the protein of interest are present in the non-interacting regions with PB change. (iii) Literature survey: Relevant literature of the crystal structures was studied to check for any previously known information about these observed PB changes for each protein. The information gathered from the above sources is listed in Table [3](#T3){ref-type="table"}. The B-factor distribution for the non-interacting residues with structural change varied from 'low' (normalized B-factor \< −1, see Methods) to 'very high' (normalized B-factor \>3, see Methods) values. Unfortunately, PDB SITE records and CSA did not provide information in most cases. Literature survey, although unable to account for all the structural changes observed in the non-interfacial region, indicates that many of these changes are allosteric (15/34) \[[@B56]-[@B70]\].
######
Features of non-interacting regions with substantial conformational change upon protein-protein interaction
-------------------------------------------------------------- --------------------------------------------------------------------- ---------------------------------------------- ------------------------------------------------------ --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
**Protein** **B-factor** **PDB SITE record/Crystal atlas** **Crystal packing**
Bound Unbound
*Proteins with almost no structural change at the interface*
PIP3 kinase (O) Poor resolution and many missing residues NI \-
HISF protein (O) *H* *H* Catalytic site residues not present in NIR_PBc/IR \-
*UCH-L3 (O)* *Nh* *Nh* *Out of 4 active site residues, one in IR* *NIR_PBc is near IR and are rigid in unbound form, one of the symmetry-related molecules is situated \~6Å towards this region.*
Son of Sevenless (O) H Nh NI \-
TGF-beta (O) Missing residues near this region NI \-
HPr kinase C-ter domain (E) Resolution of PDBs is 2.8 Å NI \-
Cystatin (E) *H* *H* NI \-
DH/PH domain of TRIO (O) Temperature factors for region under consideration abnormally high! NI \-
*Actin (O)* *H* *Not so high* *4/12 important residues are present in NIR_PBc* *Region involved in dimerization in unbound form*
*Alpha-1-antitrypsin (E)* *H* *H* *Important residues present in IR* *-*
*TGFbeta receptor (O)* *Nh* *Nh* *3/5 of active site residues are present in NIR_PBc* *-*
*Vitamin D binding protein (O)* *Nh* *Nh* *NI* *NIR_PBc values are close to the ones observed are IR and rigid in unbound form, one of the 4 symmetry-related molecules comes \~7Å towards this region. Another NIR_PBc value is far away from IR and is slightly mobile - the same symmetry-related molecule comes \<5Å close to this region!*
*TolB (O)* *Nh* *H* *NI* *-*
Ran GTPase (O) Nh (nearby IR), H H, Nh NI \-
Sporulation response factor B (O) Missing residues near this region NI *-*
Ran GTPase (O) H Nh One important residue present in IR \-
*HEW lysozyme (A)* *Nh* *Nh* *Catalytic site residues not present in NIR_PBc/IR* *-*
Transferrin receptor ectodoma in (O) Poor resolution. Important residues not present in IR/NIR_PBc *-*
Anthrax toxin receptor (O) Missing residues near this region NI *-*
Xylanase inhibitor (E) *H* *H* Catalytic site residues not present in NIR_PBc/IR \-
Fab (A) *H* *Very high* NI \-
*Complement C3 (O)* *Nh* *Nh* *Important residues are not present in NIR_PBc/IR* *-*
*Proteins with substantial structural change at interface*
*Arf1 GTPase (O)* *Nh* *Nh* *-* *-*
*Ran GTPase (O)* *Nh* *Nh* *Important residues are present in NIR_PBC/IR* *-*
-------------------------------------------------------------- --------------------------------------------------------------------- ---------------------------------------------- ------------------------------------------------------ --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
We observed that most of the proteins with changes are signalling proteins (17/22 in Case 1: Conformational changes in non-interacting surface regions of proteins with invariant interfaces & 8/12 in Case 2: Conformational changes in non-interacting surface regions of proteins with altered interfaces). Therefore, 25/40 'other' complexes (predominantly signalling proteins) from the PPC dataset show significant structural changes at distal sites, indicating prevalence of this phenomenon in signalling proteins. In contrast, only 7/25 and 3/11 complexes of the enzyme-inhibitor and antigen-antibody classes, respectively, show such changes.
Detailed information about the residue positions and the nature of conformational change observed in the residues possibly forming the target site for all examples of Case 1 and Case 2 are listed in Additional file [10](#S10){ref-type="supplementary-material"}: Table S3 & Additional file [11](#S11){ref-type="supplementary-material"}: Table S4, respectively.
Although literature studies implicate allosteric communication to be the reason for the observed structural changes away from the interface in nearly half of the complexes, we did not get clues for the other cases. Since flexibility of a region is known to be good indicator of functional relevance \[[@B45]\], we used this as a metric to identify the biological relevance of the structural changes in all the cases. Coarse-grained normal mode analysis (NMA) \[[@B71]\] is an effective and widely used method to identify intrinsic dynamics of biomolecules at equilibrium conditions solely based on their 3-D structures. This approach computes all possible vibrational modes in which the molecule can move. Studies show that biologically important functional motions are almost always captured within one or many low-frequency modes, since they require the least energy for conformational transitions \[[@B72]\]. Each mode indicates an intrinsic tendency for collective reconfiguration at particular regions. Coarse grained NMA has been applied to various aspects of structural biology, ranging from prediction of functionally relevant motion in proteins and assemblies, refinement of cryo-EM structures, identification of notable evolutionarily conserved dynamic patterns in protein families, to guiding protein docking to proceed along trajectories deemed to be functionally relevant \[[@B72],[@B73]\]. Specifically, a study of four protein-protein complexes using different variations of NMA to identify the regions and directionality of structural change revealed that these changes correlate with intrinsic motions of the protein in the unbound form \[[@B74]\]. A Gaussian network model (GNM) --based NMA of the unbound proteins in Case 1 and Case 2 sets that contain 'non-interacting regions with PB change' with low B-factors in both unbound and bound forms were carried out using oGNM web server, to identify regions exhibiting intrinsic motion, which has largely been observed to correlate with biologically relevant regions \[[@B75]\]. A summary of the normal mode analysis results are presented in Table [4](#T4){ref-type="table"}. NMA indicates that structural changes away from the interface in some of the complexes are functionally relevant. A study by Chennubhotla and Bahar indicates that a method that combines information theoretic concepts with normal mode analysis can be used to determine the communication mechanisms encoded in the structural topology of the protein \[[@B76]\]. Based on these studies, allostery, which has already been observed in 15/34 complexes according to literature reports, appears to be the most likely mechanism to explain the structural changes occurring at regions away from the interface arising from protein binding.
######
Results of normal mode analysis for proteins with 'non-interacting regions with PB change'
**PDBcode** **Region of interest** **Mobile/Rigid?** **Whole/Embedded** **Mode**
---------------------------------------------------------------------------------------------------------------------- ------------------------------------ ------------------- -------------------- ----------
*Cases with substantial change in non-interacting residues whereas interacting region has very less change*
2HQS:A Interface region Highly mobile Whole 5
NIR_PBc far away (mainly 86-91) Highly mobile Whole 3
1GHQ:A Interface region Highly mobile Whole 1
NIR_PBc far away (264-274) Partially mobile Embedded 1
1OPH:A Interface region Highly mobile Whole 1
NIR_PBc far away (120-123) Mobile Embedded 1
NIR_PBc near interface (191-194) Mobile Embedded 1
1KXP:D Interface region Rigid \- \-
NIR_PBc far away (314-325) Low mobility Embedded 1
*NIR_PBc near interface (250-258)* *Rigid* *-* *-*
2BTF:A Interface region Rigid \- \-
*NIR_PBc far away (156-158)* *Rigid* *-* *-*
1XD3:A Interface region Partly mobile Embedded 2
*NIR_PBc near interface (89-93)* *Rigid* *-* *-*
1BVK:F Interface region Partly mobile Whole 1
NIR_PBc with interface (100-105) Mobile Embedded 2
1B6C:B Interface region Rigid \- \-
NIR_PBc near interface (15-22) Mobile Whole 3
NIR_PBc near interface (152-157) Mobile Embedded 4
*Cases with substantial change in non-interacting residues near interacting region as well as in interacting region*
1I2M:A Interface region Rigid \- \-
NIR_PBc far away (145-149) Mobile Embedded 4
NIR_PBc near interface (32-29) Partly mobile Embedded 2
\*The abbreviations used are: *IR* - Interface regions, *NIR_PBc* -- Non-interacting regions with PB change.
This table lists the results of the analysis using NMA regarding the intrinsic tendency for regions of interest to be rigid or mobile. The term 'Embedded' indicates that the region under consideration possesses intrinsic mobility as part of a segment, whereas the term 'Whole' indicates that the region is independent. NIR_PBc which are indicated to be mobile are shown in bold formatting and those indicated to be rigid are italicized. Note that the PDB codes and chains provided are for the bound form whereas the corresponding unbound form structure was used for NMA.
Apart from NMA, the extent of evolutionary conservation of these regions was also determined, using Jensen-Shannon divergence measure. For regions where normal mode analysis did not provide an indication of intrinsic motion, we analyzed whether crystal packing effects could provide an explanation for conformational changes. To determine this, symmetry-related molecules were generated for the bound and unbound molecule using PyMOL \[[@B77]\], and checked to find any crystal packing which could cause the change. A discussion of a few specific cases is presented below.
a\) TolB -- Peptidoglycan associated lipoprotein (Pal) complex. The proteins TolB and Pal constitute the complex used by *Escherichia coli* and other 'group A colicins' to penetrate and kill cells \[[@B78]\]. TolB protein comprises of two domains -- a smaller N-terminal domain and a larger C-terminal β-propeller domain which interacts with Pal protein. Large rigid-body motions and conformational changes were seen far away from the interface between the unbound and bound forms of the TolB protein (Figure [6](#F6){ref-type="fig"}A, encircled region). The biological relevance of these changes is supported by experiments which prove that Pal binding results in conformational changes being transmitted to N terminal α/β domain of TolB \[[@B79]\]. Further, a recent study shows that TolA binds to the N-terminal region exhibiting structural change in TolB \[[@B80]\], indicating that the structural changes occurring upon Pal binding serves as an allosteric signal. Additional support comes from GNM-based normal mode analysis of the unbound form. The region of our interest (mainly residues 86--91), was seen to have intrinsic tendency for reconfiguration in the second most significant mode pertaining to local motions (mode 3, Figure [6](#F6){ref-type="fig"}A, purple coloured region). However, we identify that this region is not evolutionary conserved and most of the sites occurred in the least conserved bin of residues in the protein.
{#F6}
b\) Complement C3 and Epstein-Barr virus receptor C2 complex. Complement component C3d binds to antigenic molecules. This binding helps in further amplification of B cell responses as a consequence of the simultaneous binding of antigen-bound C3d with complement receptor type 2 and binding to B cell receptor via bound antigen \[[@B81]\]. Complement C3's interaction with C2 receptor causes conformational changes, identified using PBs, at residues 264--274. No literature information specific for this region was available. However, GNM analysis of the unbound form indicated that the region near to interface and the region in the opposite side have intrinsic motion (Figure [6](#F6){ref-type="fig"}B, purple coloured regions). The region of our interest occurs opposite to the interface and is indicated to be partly flexible, implying that this motion could be biologically relevant (Figure [6](#F6){ref-type="fig"}B, green coloured segment in purple coloured region). However, this region is moderately conserved.
c\) Ran GTPase and Regulator of chromosome condensation (RCC1) complex. Ran GTPase is a key component of G-protein signaling. It serves as a molecular switch which cycles between GDP- and GTP- bound states. It requires regulators for enhancing its low intrinsic hydrolysis and nucleotide dissociation rates. Guanine nucleotide exchange factors form the latter group which bind to G-proteins and induce rapid dissociation of bound GTP and hence enable fast activation to GTP bound form. The structure under consideration is a complex of Ran GTPase with the guanine nucleotide exchange factor RCC1 \[[@B82]\].
Ran GTPase, which adopts the P-loop containing nucleoside triphosphate hydrolases fold, contains two regions of interest, one near the interface (residues 41--44) and the other far away from the interface (residues 151--155) (Figure [6](#F6){ref-type="fig"}C, encircled regions). The interface also undergoes substantial changes on binding. The region of interest near the interface is seen to be intrinsically mobile in the second most important mode. The region of interest far away from the interface is seen to be a part of a mobile region in the fourth important mode. This region is very near to GDP-binding site in the unbound form. It appears that binding of Ran GTPase and RCC1 causes structural changes at distant sites (GDP-binding site) to bring about exchange of nucleotides. This case provided a clear example of signal transduction within the molecule to bring about the desired biochemical effect. Such sites can also probably be targeted by human intervention to prevent disease manifestations, such as cancer in Ran signalling pathway \[[@B83]\]. The sampling of homologous sequences was not diverse enough to obtain a reliable answer about its evolutionary conservation.
Structural changes away from the interface observed largely in proteins with structurally altered interfaces
------------------------------------------------------------------------------------------------------------
Since structural changes away from the interface appear to be common on protein binding (*n* = 34/76), we studied the structures of the proteins in the complex to understand if there are any common characteristics of the protein exhibiting these changes (target protein) versus the binding partner (effector protein).
When we analysed the type of interface (pre-made, induced-fit, other) in the target protein, we observe that pre-made interfaces constitute only 7/34 complexes undergoing structural changes away from the interface. Induced-fit and moderately induced-fit ('other' interfaces) constitute the bulk, accounting for 15/34 and 13/34 complexes, respectively. Allostery appears to be the most plausible explanation to connect protein binding with structural changes away from the interface, as discussed in the previous section. Therefore, most of the proteins comprising of pre-made interfaces (26/33) do not appear to be the target proteins involved in allosteric communication. On the contrary, they seem to serve as the effector molecules for transmitting the allosteric signal to the partner protein. Comparison of the global RMSDs of case 1 complexes indicates that the target proteins (Cα RMSD = 1.52 ± 0.71 Å) are significantly more flexible than their effector proteins (Cα RMSD of 0.86 ± 0.60 Å with a *p* value of 0.0018 for a Wilcoxon paired test). However, the disparity in the values could be partly caused due to the differences in lengths (effector - 183 ± 91, target - 389 ± 247, Wilcoxon paired test *p* value - 0.0006) since RMSD is dependent on the number of residues. For proteins of case 2 complexes, there is no significant variation in terms of their lengths (Effector proteins -- 236 ± 105, Target proteins -- 245 ± 88, Wilcoxon paired test p value -- 0.73). However, the global RMSDs of the target proteins (1.65 ± 0.64) are slightly yet significantly higher than those of the effector proteins (1.04 ± 0.67) (p value -- 0.0342, Wilcoxon paired test). Therefore, it appears that binding of an effector protein causes changes at the interface of the target protein that are propagated towards the distant allosteric site, providing credence to the views implicating flexibility in allosteric modulation \[[@B84]\].
Discussion
==========
Availability of bound and unbound structures of proteins provides an opportunity to address various questions regarding structural alterations occurring due to protein-protein interactions. Our study underlines that macromolecular liganded forms of proteins undergo larger structural alterations in terms of change in local conformation (captured using PBs) as well as atomic positions (captured using RMSD) compared to unliganded proteins (Figure [1](#F1){ref-type="fig"}). These changes are much larger than those observed due to random fluctuations characteristic of intrinsic flexibility or experimental artifacts (see Additional file [3](#S3){ref-type="supplementary-material"}: Figure S1).
Non-obligatory complexes occupy a niche position as key regulators of cellular homeostasis. Their specific and timely association and dissociation are crucial for bringing about required biological function. Spatial and temporal regulation of the interacting proteins is one of the ways of avoiding unsuitable complexation \[[@B85]\]. The other mechanism could be the use of different conformations of binding sites, which provide favourable or unfavourable binding-competence to the partner.
Transformation of binding site structure into the active form can serve as a switch to ensure correct binding at the appropriate time. Our analysis of structural alterations provides credence to this view. Surprisingly, pre-made interfaces, which are structurally invariant upon binding, shows distribution of %PB changes similar to that observed for induced-fit interfaces. This indicates that there is some extent of conformational change in all interfaces; only the nature and magnitude varies (Figure [2](#F2){ref-type="fig"}B). Additionally, interface of the partner of pre-made interface is usually observed to undergo significant structural changes (Figure [2](#F2){ref-type="fig"}A). In essence, there are no 'completely pre-made' interfaces in non-obligatory complexes. Crucially, significant structural changes are observed at backbone level in most of the interfaces used in this study. It is well known that side-chains undergo large structural changes upon protein-protein complexation \[[@B86]\]. Considered together, these results support the view that structural conformations by themselves can serve as a good mechanism to implement the required tight regulation. Lower magnitude of structural changes is generally observed to optimize complex formation, whereas larger magnitude of structural changes is observed to remove steric clashes.
We also observe a substantial proportion of instances with significant conformational changes in non-interacting regions away from the interface (Figure [4](#F4){ref-type="fig"}). Identification of these cases is facilitated by the ability of PBs to capture subtle structural variations. Observation of structural changes away from interface changes has been reported previously \[[@B87]-[@B89]\]. They could be due to various factors:
1\. Flexible regions are dynamic and can take up several distinct conformations, which can have specific functional relevance \[[@B45]\]. Several studies have revealed that flexibility is localized to certain regions of protein structure and such dynamic sites are usually involved in both small and large molecular interaction \[[@B90],[@B91]\] and enzymatic catalysis. In our study, the interface regions of TolB - Pal complex (Figure [6](#F6){ref-type="fig"}A) and Complement C3 - Epstein-Barr virus receptor C2 complex (Figure [6](#F6){ref-type="fig"}B) are shown by normal mode analysis to be intrinsically mobile.
2\. Studies show that thermodynamic entropy redistribution is a common outcome of protein-protein interaction, irrespective of the net change in entropy after complexation \[[@B92]\]. Loss of entropy at interacting sites is many times accompanied by gain of entropy in other regions of surface. 'Entropy-entropy compensation' may be due to significant intermolecular motion between the interacting molecules, which recovers about half of the entropy lost due to rotational and translational components \[[@B92]\]. This compensatory mechanism has been postulated to be the mechanism responsible for high-specificity binding of multiple ligands at the same region of a protein.
3\. The region may be functionally relevant, for e.g. a ligand/macromolecule binding site, whose conformation is regulated by an allosteric mechanism. Since binding sites are observed to be a combination of flexible and rigid sites \[[@B90]\], the signal based on protein-protein complexation may alter the stability and facilitate conformational change at the functionally relevant distant region. The complex of Ran GTPase with its cognate guanine nucleotide exchange factor probably utilizes this mechanism since complexation helps in altering the accessibility to the ligand on Rho protein (Figure [6](#F6){ref-type="fig"}C).
4\. Crystallization is known to induce substantially altered conformations \[[@B44]\]. In our study, we ensure that this bias is accounted for (Table [3](#T3){ref-type="table"}) and that the conformational changes observed are not due to such effects.
5\. Trivial factors, such as missing residues near the region of interest (or) the region being near termini, could contribute to such changes \[[@B93]\]. Since we ruled out complexes exhibiting such changes (Table [3](#T3){ref-type="table"}), the changes observed have other biological origin.
In-depth analysis of several complexes using rigorous coarse-grained NMA and literature survey indicates that a fair proportion of structural changes upon protein-protein complexation are allosteric (Figure [6](#F6){ref-type="fig"}, Tables [4](#T4){ref-type="table"}). Such communication is largely enriched in signalling proteins, which seems plausible considering the complex regulation of signal transduction pathways achieved using the interplay of several modular elements \[[@B12]\]. The lesser frequency of occurrence of such changes in enzyme-inhibitor and antibody-antigen complexes is expected. In the case of the former, their interaction is usually the result of an allosteric modulation and in the latter, a very high-affinity complex is formed, which needs to be cleared.
The classical view of allostery is as a mechanism of effector binding causing functionally relevant conformational changes at a distant site \[[@B10]\]. The salient features of the models involves two key attributes: the presence of two conformational states of the protein, one stabilised in the unbound state and the other favoured upon binding of the allosteric effector, and induction of structural change at the target site leading to functional modulation. However, studies in the last two decades have thrown new light on this phenomenon. The observations of allosteric modulation in the absence of conformational change \[[@B94]\] and the introduction of allosteric perturbation in non-allosteric proteins \[[@B95]\] have raised the viewpoint that all dynamic proteins are possibly allosteric \[[@B96]\]. These studies indicate that proteins in their unbound states exist in several conformational sub-states, characterized by different population densities \[[@B97]\]. Allosteric perturbation results in change in the relative populations of these conformers \[[@B98]\]. Such studies resulted in a paradigm shift in the understanding of allostery from a structure-centric to a thermodynamics-centric phenomenon \[[@B98]\]. Although newer studies on allostery indicate that change in dynamics also enables allosteric communication in many cases \[[@B94]-[@B96],[@B98],[@B99]\], in this study we have confined ourselves to the study of allostery in the classical sense, as communicated by structural changes. Surprisingly, allosteric communication established only via structural changes appears to be established in almost half of the complexes upon protein binding. Consideration of dynamics along with structural changes would most probably lead to uncovering of many more protein-induced allosteric changes. Therefore, our study suggests that protein-protein binding in the case of signalling complexes, is often likely to result in downstream effects. The smaller of the two proteins in a complex, usually comprising of an unaltered interface upon protein-protein complexation, appears to be the effector molecule in most cases. The binding event generally causes changes at the interface and concomitant structural changes at the target site.
Signalling proteins are key drug targets and the usage of allosteric modulators as drugs is gaining acceptance \[[@B100],[@B101]\]. In such a scenario, the understanding that most protein-protein interactions in signalling proteins are allosteric provides impetus for the design of allosteric modulators as drugs. Allosteric regulators provide certain advantages over traditional drugs, which are usually competitive inhibitors. Binding of an allosteric drug at a distant site provides reduced side-effects, saturability, modulation in the presence of true agonist etc. \[[@B84],[@B100]\]. We hope that knowledge of possible allosterically modified sites identified in the signalling complexes studied in our analysis (see Additional file [10](#S10){ref-type="supplementary-material"}: Table S3 & Additional file [11](#S11){ref-type="supplementary-material"}: Table S4) serves as a starting point for combating disease manifestations.
Conclusions
===========
Comparison of bound and unbound structures of protein-protein complexes enables us to address various questions regarding structural alterations occurring due to interaction. Non-obligatory complexes occupy a niche position as key regulators of cellular homeostasis with appropriate and timely association and dissociation which are crucial for eliciting the necessary biological function. Structural alterations in most of the interfaces of these non-obligatory complexes support the view that conformational features by themselves can serve as a good mechanism to implement the required tight regulation.
The interface is the most altered region in the entire protein structure upon protein-protein binding, as expected. The modifications are largely conformational in nature. In the rare case of one the partners remaining unaltered, the other partner is usually observed to undergo significant structural modification, thereby supporting the 'induced fit hypothesis' \[[@B52]\] more than the 'lock and key hypothesis' \[[@B51]\].
The observation of a substantial proportion of instances with significant structural changes in non-interacting regions away from the interface implies that the binding is likely to result in downstream effects. In-depth analysis of several complexes using rigorous coarse-grained NMA and literature survey indicates that these changes have functional relevance, with most of them being allosteric. The observation of allostery-like structural changes in about half of the transient complexes suggests this phenomenon is much more prevalent in signalling complexes than appreciated before. It also appears that the reversible nature of protein-protein association and dissociation, characteristic of transient complexes, affords nature with an attractive means to bring about allostery which is generally a reversible process.
Methods
=======
Datasets used
-------------
Two kinds of control datasets are used.
a\) *Rigid-proteins dataset (Control dataset 1):* A dataset of 50 independently determined structures of two rigid proteins (see Additional file [1](#S1){ref-type="supplementary-material"}: Table S1), bovine ribonuclease (32 structures) and sperm whale myoglobin (18 structures), were taken from Rashin et.al \[[@B44]\]. Values calculated from this dataset for different parameters are used as thresholds to account for positional coordinate uncertainty.
b\) *Monomeric-proteins dataset (Control dataset 2):* To get a general idea about the flexibility in atomic positions for a random dataset, the PDB was mined for crystal structures of proteins with the following criteria: a single chain is present in the asymmetric unit and biological unit; crystallographic resolution of the structure should be 2.5 Å or better and the structure should not contain DNA, RNA, DNA-RNA hybrid, or other ligands bound to the protein. These molecules were clustered at a sequence identity of 95% and length coverage of 100% using BLASTCLUST (<http://www.csc.fi/english/research/sciences/bioscience/programs/blast/blastclust>). Finally, the clusters were refined to contain only one entry for each PubmedID per cluster, which ensures that mutants are not considered, to arrive at a dataset containing 95 clusters (see Additional file [2](#S2){ref-type="supplementary-material"}: Table S2) of 319 independently solved protein structures.
### Protein-protein complex (PPC) dataset
The set of curated non-obligatory protein-protein interaction complexes solved in both unbound and bound form is taken from Benchmark 3.0 dataset \[[@B34]\]. The set was further pruned using PISA \[[@B102]\] and PDB biological unit information to exclude cases containing different non-biological oligomeric forms of a protein in the unbound and bound forms (eg. X-X in unbound form and X-Y in bound form) and bound to other small ligands or peptides. All antibody-antigen complexes in the original dataset in which only the bound structure of the antibody was solved were discarded since the corresponding unbound form was not available. The final dataset consists of 76 non-obligatory complexes (see Additional file [2](#S2){ref-type="supplementary-material"}: Table S2). The resolution of these entries is 3.5 Å or better. Proteins in every interacting pair in the dataset is non-redundant at the level of SCOP family \[[@B47]\]. Although a much larger dataset can be compiled if only one of the interacting proteins is available in unbound and bound form, such a dataset was not used since our objective is to compare the changes occurring in both the proteins upon complexation.
Although our dataset is intended to contain entries of identical proteins or protein domains available in both protein-bound and free forms, practically there could be some differences in the length and region of known 3-D structures in the bound and free forms. However the overwhelming majority of the same protein available in bound and free forms have \>90% sequence identity (see Additional file [2](#S2){ref-type="supplementary-material"}: Table S2) indicating that the bound and unbound forms are almost the same. In all the cases with % sequence identity less than 90%, it is observed that the aligned region is identical or contains very few substitutions. Further, of the 3 cases showing large length variation between the bound and unbound forms (PDB codes: 1gcq, 1qa9, 1e6j) only 1e6j features in our analysis of cases showing structural changes away from the interface. So, it appears that the analysis is robust to length variations between bound and unbound forms of a protein.
As mentioned before the dataset used in the present analysis was derived from the robust list of protein-protein complexes proposed by Weng and coworkers \[[@B34]\] in their protein-protein docking benchmark version 3.0. In this dataset the authors have carefully avoided the complexes with significant extent of disordered regions. Indeed in the dataset used in the current analysis none of the complex structures used has any disordered residue at the protein-protein interfaces. This could be ensured on the basis of information on missing residues given in the PDB file, by checking the distance between Cα atoms of putative adjacent residues and by checking for the presence of all the expected atoms in a residue.
Identification of interfacial residues
--------------------------------------
If the distance between any two atoms of residues from the two proteins is less than sum of their van der Waals radii + 0.5 Å, the two residues are considered to be in the interface \[[@B53]\]. The van der Waals radii were taken from the literature \[[@B103]\].
Classification of residues based on solvent accessibility
---------------------------------------------------------
The residues in a structure are classified on the basis of their residue surface accessibility (RSA) which is calculated using NACCESS \[[@B104],[@B105]\]. This parameter provides a normalized measure of the accessible surface area of any residue in the protein, calculated with respect to the extended form of the residue, using the NACCESS program. The cut-offs employed are: ≤5% RSA (buried residues) and ≥10% RSA (surface residues). The 5% cut-off was adopted from \[[@B106]\], who optimized and used it to define residues buried in monomeric proteins. Buried, surface, and interface residues constitute \~25%, 75% and 10-20% of the residues in a protein, respectively.
Quantification of structural change
-----------------------------------
Structural change is estimated for a given residue in unbound and bound forms. A sequence alignment of the unbound and bound forms performed using CLUSTALW \[[@B107]\] provides the residue equivalences. Structural change is captured using two measures: RMSD and Protein Blocks. Structural change is classically captured by means of root mean square deviation (RMSD), where RMSD is calculated as follows: $RMSD = \sqrt{1/N\sum di^{2}}$ for *i* ranging from residue 1 to *n* of the dataset and *d* is the distance between N pairs of equivalent atoms. Two measures of RMSD have been employed: Cα RMSD and all-atom RMSD, based on deviation between the Cα positions of the same residue in unbound and bound forms in the former and between all-atoms of the same residue in unbound and bound forms for the latter. Deviation in side chain positions are generally expected \[[@B86]\] whereas large backbone changes are comparatively uncommon. Therefore, the deviation between the Cα positions of the same residue in unbound and bound form is used as an indicator of structural change mainly. The changes are captured at structural level and averaged out for the entire protein or a set of residues in a protein (for e.g. interface residues) and the averaged measures are used in the analysis. Small yet significant changes in local conformation of a protein can be captured using Protein Blocks \[[@B48]\]. The three dimensional structural information in the bound and unbound forms is represented in a one-dimensional form using Protein Blocks (PBs). They consist of 16 structural prototypes, each of which approximates the backbone of a five-residue peptide. Given a 3D structure, each overlapping sequence of 5-residue fragments is associated with its closest PB. The sequence of PBs is annotated in the sequence alignment obtained using CLUSTALW. Two parameters are calculated using this measure. The first parameter indicates the presence of conformational change and is calculated as % changes in PBs between unbound and bound form (PBc). The second parameter indicates the magnitude of observed change and is calculated using PB substitution score (PBSSc) for the equivalent residues.
'Pre-made' versus 'induced-fit' interfaces
------------------------------------------
An interface with ≥0.5 Å Cα RMSD difference between the bound and unbound forms is classified as 'pre-made' interface whereas an interface with ≥1.5 Å Cα RMSD difference between the bound and unbound forms is classified as an 'induced-fit' interface. However, there are some interfaces with lower difference in terms of magnitude but with substantial difference at the interface in comparison to the rest of the surface residues (≥10% RSA). This cut-off was chosen since 90% of the interface residues have an RSA equal to or greater than this value in the unbound form. A normalization-based metric was used to identify induced-fit interfaces exhibiting smaller structural changes as $N = \frac{C\alpha RMSD_{I}}{C\alpha RMSD_{\mathit{ROS}}}$ where 'CαRMSD~I~' indicates the average Cα RMSD difference between bound and unbound form for interface, and 'CαRMSD~ROS~' indicates the average Cα RMSD difference between bound and unbound form for the rest of the surface.
For example, a value of two indicates a doubled change in magnitude of the interface with respect to the rest of the surface. This value is used as a cut-off to identify substantial changes localized to the interface.
Identification of proteins with substantial structural change in non-interacting regions
----------------------------------------------------------------------------------------
To identify cases where the interface is largely invariant/moderately altered:
*Criterion 1:* Average Cα RMSD of 'non-interacting residues with PB change' should be ≥1 Å than the average Cα RMSD of the interface residues. For this comparison, the individual segments under consideration, were superimposed using SUPER (B.S.Neela, unpublished).
*Criterion 2:* Normalized PB substitution score of 'non-interacting residues with PB change' should be ≤ −2 than the normalized PB substitution score of interacting residues.
To identify cases where there are large changes at interface:
*Criterion 1:* Average Cα RMSD of both 'non-interacting residues with PB change' and interacting region should be ≥2 Å.
*Criterion 2:* Average Cα RMSD of both 'non-interacting residues with PB change' and interacting region should be \> \> global Cα RMSD.
Analysis of B-factors
---------------------
The B-factor (temperature factor/atomic displacement factor) of an atom reflects the degree of isotropic smearing of electron density around its center \[[@B108]\]. A low B-factor indicates small uncertainty in the position of an atom. A high B-factor can be caused by different factors: high thermal fluctuations, alternate conformation of an atom, and domain motion, to name a few.
To ascertain the flexibility/rigidity of a particular residue in a structure, its normalized backbone B-factor was considered \[[@B109]\]. Normalization with respect to all the other residues provides an idea of increase/decrease in flexibility on a standard scale. Only surface residues (≥10% RSA) were considered for the normalization since all interacting residues and non-interacting surface residues form the crux of this study. The three most N-terminal and C-terminal surface residues were excluded since their B-factors are usually high and can affect the 'mean' of the values. B-factors of only backbone atoms were considered as we are studying backbone changes and also since side chain are generally more flexible than backbone atoms. The normalized B-factor per residue (*B*~*i,N*~) was computed as $B_{i,N} = \frac{B_{i} - < B_{i} >}{\sigma_{\mathit{Bi}}}$ where *B*~*i*~ is the B-factor of residue *i*, \<*B*~*i*~\> is the mean B-factor of the protein surface residues and σ ~*Bi*~ is the s.d. for the same.
Residues with backbone B-factors ≥3, ≥2, and \<1 standard deviations from the mean backbone B-factors for surface residues can be considered to have 'very high', 'high' and 'low' flexibility, respectively.
Identification of regions of protein structure with intrinsic collective motions
--------------------------------------------------------------------------------
GNM-based NMA of the unbound form of a protein was undertaken to identify intrinsic collective motions of the molecule. In this model, the biomolecule is modelled as a harmonic oscillator with every residue represented as a single site, connected by springs to nearby residues \[[@B73]\]. The oGNM web server \[[@B75]\] calculates low frequency normal modes for the unbound structure based on GNM. In GNM, the motions are isotropic by definition, thereby predicting only regions exhibiting intrinsic motion and magnitude of change. The directionality of motion cannot be predicted using GNM models. The server constructs the elastic network model of the structure by considering each of the Cα atoms as a node and identifying all interacting nodes using a distance cut-off of 10 Å. The six most low frequency modes were analyzed to check whether any of the 'non-interacting regions with PB change' far away from the interface show probable biologically relevant intrinsic motion.
Determining the extent of conservation of non-interacting residues with conformational changes
----------------------------------------------------------------------------------------------
xBased on the assumption that evolutionary conservation of a site in a protein family is an indicator of its functional relevance and/or structural integrity, the degree of conservation of all sites in a protein family was calculated using the Jensen-Shannon divergence measure. This metric operates on the premise that most sites in a protein family are not under any evolutionary pressure and hence have a distribution similar to background amino acid distribution. Sites under evolutionary pressure, such as functional or stabilizing sites, show amino acid distribution significantly different from the background distribution.
Homologous sequences for every protein in our PPC dataset were identified by a search employing PSI-BLAST \[[@B110]\] against the UNIREF90 \[[@B111]\] database at an e-value cutoff of 0.0001 for 3 iterations. Further, only sequences with ≥30% identity were considered. A multiple sequence alignment (MSA) of the query sequence with only the aligned regions of the homologous sequences was generated using CLUSTALW. The conservation scores for every site in the MSA was calculated using Jensen-Shannon divergence measure \[[@B112]\]. The sites with top 30% conservation scores are considered to be well conserved \[[@B112]\] and sites with bottom 30% conservation scores are considered to be poorly conserved.
Generation of symmetry-related molecules using PyMOL
----------------------------------------------------
Symmetry-related molecules were generated for the bound and unbound molecule using PyMOL \[[@B77]\]. The crystal packing after generation of symmetry-related molecules was checked to ascertain if any crystal packing could cause the observed structural changes in a complex.
Competing interests
===================
The authors declare that they have no competing interests.
Authors' contributions
======================
NS and AdB conceived and designed the experiments. LSS and SM performed the experiments. LSS, SM, AdB and NS analyzed the data and wrote the paper. All authors read and approved the final manuscript.
Supplementary Material
======================
###### Additional file 1
**Table S1.** List of PDB codes of structures of control datasets used in this analysis.
######
Click here for file
###### Additional file 2
**Table S2.** Details of various features of PPC dataset.
######
Click here for file
###### Additional file 3
**Figure S1.** Distribution of parameters capturing structural change for Control and Test datasets.
######
Click here for file
###### Additional file 4
**Figure S2.** Distribution of parameters capturing structural change with respect to 'interface area' and 'length of protein'.
######
Click here for file
###### Additional file 5
**Figure S3.** Distribution of all-atom RMSD values for PPC dataset.
######
Click here for file
###### Additional file 6
**Figure S4.** Different kinds of interfaces.
######
Click here for file
###### Additional file 7
**Figure S5.** Parameters for identifying rigid-body movements.
######
Click here for file
###### Additional file 8
**Figure S6.** Different types of structural changes seen at interfaces.
######
Click here for file
###### Additional file 9
**Figure S7.** Distribution of parameters for interface vs. non-interacting surface regions per protein.
######
Click here for file
###### Additional file 10
**Table S3.** Details of non-interacting surface residues with PB change in Case 1 complexes.
######
Click here for file
###### Additional file 11
**Table S4.** Details of non-interacting surface residues with PB change in Case 2 complexes.
######
Click here for file
Acknowledgements
================
This research is supported by Indo-French collaborative grant (CEFIPRA/IFCPAR number 3903-E) between AdB and NS. Part of this project is also supported for NS by Department of Biotechnology (DBT). Part of this was supported for AdB by grants from the Ministry of Research, University Paris Diderot -- Paris 7, National Institute for Blood Transfusion (INTS) and the Institute for Health and Medical Research (INSERM). LSS and SM are supported by CEFIPRA/IFCPAR and DBT, respectively. The authors are grateful to Dr. Manoj Tyagi for providing the program to perform PB assignments and to Dr. Agnel P. Joseph for providing the PB substitution matrix. The authors express their immense gratitude to Professor Ruth Nussinov for key suggestions.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1-1}
============
Image-guided Percutaneous Needle Biopsy (PNB) is an interventional procedure performed by radiologists ([@ref1]-[@ref5]) with the aim to obtain cells or tissue for diagnosis by the insertion of a needle into a suspected lesion.
It's a relatively non-invasive procedure, and it has absolute advantages compared to open or excisional biopsy.
Success of PNB is related to proper patient selection, preparation and adequate procedural planning ([@ref6]-[@ref8]).
Planning and procedural phases {#sec1-2}
==============================
PNB implicates the involvement of interventional radiologists in multidisciplinary boards ([@ref9]-[@ref21]). The radiologyst has a key role in the pre-procedural phase: to evaluate potential contraindications and risks of PNB, to confirm the indications for PNB and to identify the optimal target and the selection of the proper imaging guidance.
Indications to PNB are the characterization of nature (benign or malignant) of a lesion ([@ref22]), the diagnosis and staging of a tumor, and biological or immunohistochemical/genetic analisys on tissue ([@ref7], [@ref23]).
Although PNB is a relatively non-invasive procedure, there are some contraindications, such as the alteration of coagulation status (specially if it can't be correctable) and bleeding risk, the patient's clinical status (to tolerate bleeding or anesthesia) and cooperation.
The main imaging-guide modalities are ultrasound (US) ([@ref24]-[@ref26]) and computed tomography (CT) ([@ref27]-[@ref35]); other uncommon imaging techniques are fluoroscopy, magnetic resonance imaging (MRI) ([@ref36]-[@ref47]), and positron emission tomography CT (PET-CT) ([@ref6], [@ref7], [@ref48]).
US guidance has a wide use because of portability, lack of ionizing radiation, and low operating cost. Real time imaging allows to visualize and track the needle throughout its entire pathway and is useful even in lesions moving on respiratory motion; Color-Doppler ([@ref24]) aid in vascular structures visualization. Furthermore, in selected patients, US contrast-injection increases lesion characterization on the surrounding tissue. Freehand or needle-guided technique are both suitable. Compared to the freehand technique, the guided technique is limited by a fixed angle. Limits of the US-guidance technique are the operator experience and the appropriate acoustic window view, such as the difficulty to penetrate air-filled structures and bone ([@ref49]).
Compared to US, CT has a better preprocedural planning of PNB, because of its high spatial resolution and large field of view. It permits multiplanar reformations (MPR) to obtain a more adequate path of needle. An intravenous contrast injection may be required to increase accuracy on lesion visualization.
Other imaging guidance modalities are: CT-fluoroscopy, that allows a real-time visualization of the needle, advancement reducing procedural time, but it exposes operators and patients to radiation doses; MR-guidance, despite excellent soft tissue contrast and lack of ionizing radiation, isn't currently feasible because of increased costs and procedure time, the lack in appropriate open-scanner and MRI-compatible instruments;
PNB includes two basic techniques for sample acquisition: fine needle aspiration biopsy (FNAB) and core needle biopsy (CNB) ([Figg. 1](#F1){ref-type="fig"}-[6](#F6){ref-type="fig"}) ([@ref50], [@ref51]).
{#F1}
{#F2}
{#F3}
{#F4}
{#F5}
{#F6}
FNAB device extracts individual cells for cytological evaluation using a small needle (18-25G) with inner stylet. Once in place, the stylet is removed, a syringe is attached to the needle, and cells are aspirated. Small lesions, necrotic tumors or lesions close to critical structures are its main targets. The most commonly devices used in retroperitoneal biopsies are the spinal needles and the Chiba needles ([@ref52]).
CNB devices use larger needles (9-20G) with different mechanisms (manually or automatically cutting systems) to extract a piece of tissue for complete histologic evaluation ([@ref53]).
A safe and proven technique is the use of coaxial needle: the biopsy needle is introduced coaxially into a guide needle (9-19G), previously advanced nearby the target. It doesn't increase the recurrence of complications and allow multiple specimen samples in a single puncture and decrease the tumor cells seeding risk along the needle tract ([@ref48], [@ref54]-[@ref58]).
The extracted samples are then smeared on glass slides and fixed (FNAB) or placed in formalin (CNB); for bacteriological analysis the sample is sent in saline for culture ([@ref59], [@ref60]).
Post-procedural phase {#sec1-3}
=====================
Retroperitoneal PNB is considered a minimally invasive and safe procedure.
There are major and minor complications, related to the technique (bleeding, infection, perforation, tract seeding) or to organ specific injury (such as haematuria, pneumothorax, haemoptysis, air embolism).
After the procedure and before discharge, imaging control is generally obtained and documented to detect immediate possible complications; equally, vital signs monitoring and clinical observation are required for a few hours following the procedure. In case of major complications, hospitalization in appropriate environment should be guaranteed.
Technical success of PNB varies greatly depending upon the size and location of the target, benign or malignant nature of the lesion, number of samples obtained, availability of an onsite cytopathologist, IRs' and pathologists' experience, equipment availability ([@ref61]).
Clinical success of PNB is the usefulness of the procedure in terms of improvement of patient care.
In case of non-diagnostic biopsy a repeated biopsy should be considered, such as different techniques or approaches modalities (surgical biopsy or open access) ([@ref62]-[@ref67]).
Conclusions {#sec1-4}
===========
Retroperitoneal Percutaneous Needle Biopsy is a minimally invasive, well established and safe procedure, with a low rate of complications and high diagnostic yield.
Radiologist plays a critical role in the entire management of the patient, since the procedure planning until the patient discharge.
PNB is gaining an even more crucial role, specially with the development of molecular personalized treatment, so avoiding in several patients more invasive diagnostic procedure.
Ethical approval: {#sec1-5}
=================
This article does not contain any studies with human participants performed by any of the authors.
Conflict of interest: {#sec1-6}
=====================
None to declare
| {
"pile_set_name": "PubMed Central"
} |
Permafrost temperature has increased by 1°C to 2°C in the Russian north during the last 3 decades ([@CIT0001]). Northern Alaska and northern Canada also have warmed to a comparably high level during the same period ([@CIT0002]). Recent observations indicate that this warming is highly coupled with a shortened winter season and the degradation of the ice-rich permafrost ([@CIT0003]--[@CIT0005]). In the Arctic, these impacts are profoundly affecting communities because even relatively small landscape changes due to warming have been shown to impact the fragile tundra socio-ecological systems out of scale with their spatial extent ([@CIT0005],[@CIT0006]).
Most terrestrial and aquatic components of the Russian tundra are seasonally exploited by the indigenous nomadic reindeer herders, hunters, and fishers ([@CIT0005]). Because of the high level of interdependence between the human and environmental conditions and poor access to health services in the Russian far north, rapid shifts in hydrologic cycles in the tundra are expected to have far-reaching consequences on health and the social fabric of the local indigenous communities ([@CIT0007]).
Across the Russian north, non-emergency medical care is provided through a sparse network of small, village-based health clinics. Nomadic Nenets use the traditional reindeer sled, walking, and increasingly, the snowmobile, as their means of transportation across vast distances. They migrate several hundred kilometers with large reindeer herds in the spring and autumn. The migration paths connect the summer pastures on the nutrient-rich grassy tundra zone along the Arctic coast and the winter pastures within the forest-tundra zone. During these migrations, the nomadic reindeer herders pass through the same villages that are located on the long-established annual migration routes. Non-emergency visits to the health clinic typically take place during these transmigrations, when the herder families receive maternal, preventative, and primary care. It is reasonable to assume that the appearance of new lakes, marshland, and streams as well as a rapid change in the size of the existing water bodies is likely to pose new geographic barriers that may reduce or preclude the herders' access to the already sparse, remote, and often difficult to reach preventive and health care services.
According to 1 study of Nenets herders, the main effects of the recent warming are increases in open water in the tundra and frequent, unusually warm, autumns ([@CIT0008]). These events create transportation conditions that are difficult and unpredictable ([@CIT0005]). In the summer months, when the herders are in the coastal tundra, they are the furthest from medical help. The appearance of a new stream in the previously dry valley from 1 year to another can block the most direct route to the village and its clinic and make travel there exceedingly difficult. In some cases, this could add several days of travel on new and unfamiliar terrain to an already long journey to reach medical care. Fall-through-ice events may occur along these new routes and may present additional health hazards. However, the data on these incidents was not available at the time this article was being written.
Traditionally, autumn transmigration began with the arrival of stable sub-freezing temperatures, which establish firm ice on the fresh water bodies allowing the herders and thousands of reindeer to cross the rivers and lakes. The recent rapid warming created delayed formation of stable ice and concomitant setbacks for herders in leaving remote coastal pastures and in accessing health clinics. It also may extend the herders' stay in the grassy tundra pastures on the Arctic coast, often hundreds of kilometers from the closest settlement. In the Arctic, where geographic access to kin and communities of other herders are critical for survival, ability to reach other herders, scattered across dozens of camps on the vast tundra, is also affected by the nascent hydrological and environmental changes. For example, when footpaths that once linked age-old camps of different families become inundated the herders cannot access other camps by foot or sled.
Other problems related to the recent hydrologic changes reported by Nenets herders include loss of the valuable reindeer grazing and calving habitats to newly appeared lakes and marshland, loss of the long-established campsites and traditional migratory routes, and loss of the sacred religious sites and burial grounds to inundation ([@CIT0009]). Reported problems securing food include impeded access to wild berries and mushrooms and to hunting and fishing areas; these problems may affect the nutritional status of the herders ([@CIT0008]). Securing food is critical for nomadic Nenets because reduced access to traditional food forces a greater reliance on non-traditional food that is only available in the distant villages, which in turn become less geographically accessible due to the changing, wetter landscape ([@CIT0010]).
Nomadic Nenets and reindeer husbandry in the Russian Arctic {#S0002}
===========================================================
Nenets are indigenous Arctic people living predominantly along the Arctic Ocean in northwest Russia, and many still practice reindeer herding and the nomadic way of life. The nomadic Nenets and their predecessors began hunting and harnessing reindeer during migrations between 500 and 1100 AD ([@CIT0011]); herding became progressively more intensive after 1600 AD ([@CIT0012]). Nenets retain a unique culture, which strongly emphasizes their ties to the reindeer, the tundra, and to the nomadism, with stewardship of the tundra ecosystem playing a key role. In the Nenets Autonomous Okrug (NAO), families engaged in full-time reindeer herding annually migrate up to 1,200 km with their herds between the grassy summer pastures on the coast and the winter pastures in the forest tundra south of the NAO and in the Arkhangelsk Oblast ([@CIT0008]). The NAO ([Fig. 1](#F0001){ref-type="fig"}) is home to the second largest Nenets populations in the world ([@CIT0013]). In 2008, the NAO (176,700 km^2^) was home to 5,623 Nenets, who constituted 13.5% of the total population of the Okrug ([@CIT0008]).
{#F0001}
Reindeer (*Rangifer tarandus* L.) are important nutritionally, economically, and socio-culturally within the tundra socioecological system (SES). Reindeer husbandry is the key traditional occupation and a way of life for Nenets in the NAO where the *Rangifer tarandus* population is the second largest in Russia ([@CIT0009]). There were 157,000 domesticated reindeer in the NAO in 2007, and Kanin Peninsula in the far northwestern NAO had the largest reindeer population in the Okrug ([@CIT0008]), providing the main source of protein to the herder familes, as well as skins and income from sales of surplus deer meat. The total territory of reindeer pastures is 132,000 km^2^, or 75% of NAO land area ([@CIT0008],[@CIT0009]). There are some 20 indigenous reindeer herder groups and the number of individuals employed in reindeer husbandry recently grew from 818 (2003) to 1,100 (2007). Approximately, two-thirds of the NAO\'s herders still lead nomadic lives.
Reindeer husbandry on Kanin Peninsula {#S0003}
=====================================
Our study site is Kanin Peninsula (14,673 km^2^), located in the extreme northwestern NAO---surrounded by the White Sea to the west and the Barents Sea to the north and east. Kanin is relatively flat, undulating, low-laying grassy tundra, punctuated by many small lakes and vast expenses of marshland, rivers, and streams that are deeply cut into the surface and not crossable during the summer months. The southern part of the peninsula is called Kanin Neck. This area is especially narrow and is composed of low-lying lakes and marshland, and thus is largely impassable in the summer. During transmigrations, when herders travel with thousands of reindeer and loaded sleds, they can only cross in and out of Kanin Peninsula when the open water on Kanin Neck is frozen, which creates an objective measure of the warming\'s effect on the timing of transmigration and on arrivals to the clinic.
Unlike some of the other parts of the NAO, Kanin has not been affected by the recent hydrocarbon development boom in the region and the local population remains very small, with 5 villages and the only full-time health clinic and hospital in Nes (2007 pop., 1,256). During the Soviet era, the herders were organized into Obshini (communities) with non-overlapping calving and grazing territories for their herds, and this structure is still largely in place. On Kanin Peninsula, there are 2 herder communities, Kanin Obshina (370 members) and Yamb-To Obshina (230 members) each comprising of smaller, kin-based herder groups ([@CIT0008],[@CIT0009]). The Kanin Obshina community\'s territory is along the White Sea coast of the peninsula and the Yamb-To territory is on the eastern side of the peninsula along the Barents Sea.
During the spring and fall migrations, Kanin Obshina herders pass through the village of Nes where the health clinic is located, while the Yamb-To Obshina pass through the village of Vizas. The location of Nes, on the traditional migration paths of the Kanin Obshina just south of Kanin Neck, makes it an important hub for herders---where they stop to resupply and to seek health care during spring and autumn migrations ([Fig. 1](#F0001){ref-type="fig"}).
The local health care providers and the tundra residents have reported that temperature change and shorter, warmer winters have led to the expansion of the bodies of water, which causes hitherto unseen changes in the landscape and delays migration along the traditional routes travelled by generations of reindeer herders during the annual migrations ([@CIT0008]). Since 1979, dates have been recorded for the herders' fall arrival at the Nes clinic during their transmigration south and for some of the spring arrivals during their transmigration north. These dates provide useful information on the chorology of the change in tundra conditions with respect to health care access.
The purpose of this research is to quantify the effects of temperature anomalies and of increases in open water on transmigration and health care access by Nenets reindeer herders in northern Russia.
Materials and methods {#S0004}
=====================
We hypothesized that a pronounced increase in temperature will have an impact on access to health care in the Arctic due to its effect on transportation and on other aspects of nomadic herder livelihoods. The raising temperatures are most likely to affect access due to new barriers to transmigration and to visits to the Nes health clinic, namely because of a shorter winter season and more open water. We employed clinical records on dates of arrivals in Nes for the reindeer herders during the fall migrations from 1979--2011 ([Fig. 2](#F0002){ref-type="fig"}); these were collected by one of the authors (L. Zubov). We also used a water bodies mask derived from daily surface reflectance images produced by moderate resolution imaging spectroradiometer (MODIS) from 2004--2011 in order to calculate the total territory of the open water in the tundra.
{#F0002}
We used the mean of the temperature anomalies ([@CIT0014]) for the 4 autumn migration months for quantitative analysis of the relationship of interannual autumn temperature anomalies trends and autumn arrivals to the Nes clinic.
Remotely sensed water bodies data {#S0004-S20001}
---------------------------------
The main challenge in identifying water surfaces from remotely sensed data is the high variability of their spectral signatures. The spectral property of water is determined by the electromagnetic interaction of light with the constituent components of water via absorption or scattering processes either within the water column, at the water surface, or on the bottom of the body of water ([@CIT0015]). Consequently, the water-leaving radiance detected by the sensor shows great spatial and temporal variability, which makes its reliable discrimination particularly difficult ([@CIT0016]).
In the framework of this study, we used a methodology proposed by Pekel et al. in 2012 to detect open water surfaces in near real time. The method is based on a colorimetric approach. The rationale we employed is that the perceived color in a color composition is directly determined by the shape of the spectral signature of the signal\'s target. Our novel approach permitted us to associate both land and water surfaces with unique colors. We identified a set of thresholds to detect the open water surfaces on a 10-day basis at a resolution of 250 meters.
Some false water surface detections were manifested by a low temporal frequency. These false detections resulted from the remaining perturbations after atmospheric correction and cloud removal. Indeed, over our area of interest, the MODIS standard flags were not masking all the contaminated values; these were then included in the compositing process and induced reflectance instability and consequently some false detection. In order to discard these false detections from our analysis, we considered only the water surfaces showing an occurrence (i.e. number of water detections divided by the number of available observations) above 35% on an annual basis. Indeed, because the false detections occurred randomly, both spatially and temporally with a low temporal frequency, the lower occurrence concentrated the false detections. One drawback of this approach is that temporary water surfaces also characterized by low annual occurrence were discarded. Obviously, it decreased artificially the number of water surfaces considered in our study but because the same criterion were applied for each year, and because we used these detections in a relative way, it did not significantly impact our results. For this research, the new water body mask products were created for the period 2004 to 2009 and were mapped. An operational open water detection product from 2001 to the present will be made available in the near future.
{#F0003}
Temperature anomaly data {#S0004-S20002}
------------------------
To assess changes in temperature trends, we used monthly anomalies in air temperature data at 1.0 degree spatial resolution from the Climate Anomaly Monitoring System (CAMS) monthly gridded and station data produced by the National Oceanic and Atmospheric Administration\'s Climate Prediction Center ([@CIT0016]), shown in [Fig. 3](#F0003){ref-type="fig"}. In the world of climate studies, the term "anomaly" means the difference between the value of a quantity and its climatological mean value. A "monthly anomaly" is the difference between the original monthly value of a quantity in a given month and the monthly climatological value for that month of the year. The monthly temperature anomaly equation is written as$${r^{\prime}}_{ij} = r_{ij} - \frac{1}{N_{i}}\sum\limits_{j = 1}^{N}r_{ij}$$
Here, for months *i* and years *j*, *r′* ~*ij*~ is the monthly temperature anomaly, *r* ~*ij*~ is the original monthly value, and the remainder of the equation is the calculation of the monthly temperature (which is subtracted from the original monthly values). For CAMS, the long-term average is computed from 1950 to the present.
For the Kanin area, we computed a monthly temperature anomaly average over a territory that covers the entire study area. The data were extracted from the CAMS data set using the Ingrid code. The CAMS data set is located at the International Research Institute for Climate and Society and the Lamont-Doherty Earth Observatory (IRI/LDEO) Climate Data Library at Columbia University and available at: <http://iridl.ldeo.columbia.edu> ([Fig. 4](#F0004){ref-type="fig"}).
{#F0004}
Data analyses and statistical methods {#S0004-S20003}
-------------------------------------
To assess changes over time in herd movement, a Spearman correlation between calendar year (1979--2011) and week of arrival at Nes village was computed, along with a 95% confidence interval; parametric analysis was not attempted due to sparsity of data. To assess change over time in autumn migration season (September--December) temperature for the years 1950--2011, a local regression or LOESS plot was generated that strongly suggested a relationship between calendar year and mean temperature as having two linear components: one relevant to the years 1950--1991, the other to the post-1991 period ([Fig. 4](#F0004){ref-type="fig"}). LOESS is a strategy for fitting smooth curves to empirical bivariate data for visualization and further time series analysis. The LOESS statistical procedure is a fairly direct generalization of traditional least-squares methods for data analysis.
A mixed linear model was therefore constructed, with mean temperature for September--December as the dependent variable, calendar year as the predictor, modeled as a continuous piecewise linear relationship with a slope cut point at 1991 (in other words, as a linear spline with a single predetermined knot). First order autoregressive moving-average (ARMA (1,1)) structure was used to model autocorellation. Model residuals suggested 1974 to be an outlier year; results presented here exclude data for that year. ArcGIS 10 (ESRI, Redlands, California) software was used for mapping and SAS Release 9.2 (SAS Institute, Cary, North Carolina) software was used for statistical analysis.
Results {#S0005}
=======
[Figure 4](#F0004){ref-type="fig"} displays the CAMS temperature anomalies data for Kanin for the years 1950--2012 and the LOESS plot of the autumnal anomalies for the same period. Temperature anomalies data show the recent increase in positive anomalies throughout the year. The LOESS plot strongly suggests an increase in more recent mean autumnal temperature anomalies (September--December) after 1991 as having 2 linear components: one relevant to the years 1950--1991, the other to the post-1991 period, when the temperature began to raise sharply.
Correlation of the field-collected data on the herders' fall arrival at Nes and the calendar year for the last 32 years produced r = 0.64, p-value of \<0.001, and 95% CI (0.31; 0.82). Regression analysis estimated that mean temperature anomalies during the fall migration in September--December were stochastically stationary pre-1991 (−0.0157 Cdeg/year, SE = 0.0140, 95% CI \[−0.0437, +0.0122\], p = 0.265) but have risen significantly (p \< 0.001) since then. The rate of change is estimated at 0.7 to 2.0 Cdeg/decade +0.1351 Cdeg/year, SE = 0.0328, 95% CI (+0.0694, +0.2007); test of slope change at 1991: p = 0.001.
The amount of detected open water area and the number of water bodies per year (2004--2009) have fluctuated significantly within the study area on Kanin Peninsula---from approximately 600 km^2^ to more than 800 km^2^.
This illustrates the current instability of the total water area within the migration range of the herders. Specifically, in some years (e.g. 2004), a rapid increase in open water area is followed by a sharp drop the next year, indicating rapid draining of many water bodies. More research is needed to better understand these novel biophysical processes and their future implications on the well-being of Arctic social and ecological systems.
Visual inspection of the overlays of the annual water masks with the transmigration routes in a geographic information system or GIS revealed that many newly detected water bodies lay directly on the routes. We examined one such area on Kanin Neck (shown in [Fig. 5](#F0005){ref-type="fig"}), where large interannual changes in open water were detected by our method, and noted that there---and in many low areas on Kanin---even a modest increase in open water area caused several lakes to merge into single large bodies of water, thereby creating impassible barriers to ground transportation. In [Fig. 5](#F0005){ref-type="fig"}, Lake A has increased from 4.94 km^2^ in 2005 to 15.35 km^2^ in 2006 because the new water areas that appeared in the spring of 2006 (depicted by the red pixels) appear to have connected the previously small Lake A to the larger lake system.
{#F0005}
Discussion {#S0006}
==========
Temperature anomalies on Kanin have increased by approximately 1.4°C per decade since 1991. During the same period, there has been a marked trend of delays in herders' arrivals to the Nes clinic. These changes are likely due to the lengthening of overall migration routes in order to avoid open water as well as to the necessity of waiting until later in the fall or early winter in order for the swampland and open water to freeze making the crossing possible for both humans and reindeer.
Over the last 3 decades, unusually warm autumns have progressively delayed the herders' crossing of Kanin Neck on their way to winter pastures; their arrival at the Nes clinic has shifted from October to December ([@CIT0008]) ([Fig. 2](#F0002){ref-type="fig"}). Concomitantly, until approximately 10 years ago, herders migrated later from winter pastures near the inland villages to Kanin Peninsula and arrived in Nes on April 6--10. In recent years, these spring arrivals to the clinic shifted to March 24--28 in order to avoid earlier thaws on the herders' journey to points north ([@CIT0008]). It appears that, in order to reach the summer calving areas before ice on Kanin Neck melts and renders the summer pastures unreachable, the herders shortened winter stays near the villages and stayed longer in Kanin, where they have virtually no access to non-emergency health care.
The study\'s findings corroborate our hypothesis that a relationship exists between the recently observed warming trend and the delays in the herders' arrival at the Nes clinic. Formation of new lakes and streams, expansion of the existing water bodies, and a shorter winter (when these bodies of water are crossable over ice) all present formidable barriers to travel by foot, sled, and snowmobile. Spatial and temporal overlap of these environmental changes and their combined effect on transmigration may be responsible for the recent difficulties experienced by the herders in accessing primary and preventive care. There are several practical reasons for this.
The current increase and rapid fluctuation in the size and number of water bodies renders ground travel unpredictable because the herders and their herds must travel around the new water bodies and thus traverse greater distances in order to reach the clinic. For example, in 2006, in order to travel across the area shown in [Fig. 5](#F0005){ref-type="fig"}, the herders had to either wait until stable ice was formed over the lakes or deviate from their familiar historic routes and travel with large herds around the perimeter of the entire new water body (circled). Such deviations into unfamiliar terrain would increase total travel distance and might also increase the likelihood of injuries to humans and animals. The changes in the tundra hydrology reported here and their affect on livelihoods and access have been corroborated by the reindeer herders and local health care providers. Yet, if the current warming trend continues, it is not clear how these changes will affect the Nenets communities, such as Kanin Obshina, and other circumpolar peoples in the future. For example, do temperature anomalies and open water increase in a linear fashion or will they follow a different spatial pattern? More research is required to better understand the complex interrelationships between surface temperature increase, permafrost dynamics, shortened winter period, and increases in open water. Further development of scientifically sound methods for monitoring these changes at the community scale, as well as translating the data to be used by the community stakeholders and public health authorities, are critical steps in developing adaptation strategies in order to increase community resilience, advance healthcare delivery, and improve well-being in the changing circumpolar north.
Conclusions {#S0007}
===========
In this study, we used a novel approach to measure the relationship between temperature anomalies and access to a health care clinic by an indigenous population by combining clinician observations (arrival dates at the Nes clinic), CAM temperature anomalies, and MODIS-derived detection of open water. Until very recently, our ability to measure hydrologic change due to warming and to study its effects on to the health of Arctic peoples was severely limited by the dearth of high quality spatial and temporal data suitable for time-series analysis. Our novel use of clinic visit data, temperature records, and remotely sensed imagery permitted us to report on the recent interannual temporal shift in the herder families' visits to the Nes health care clinic. It also demonstrated a relationship between the annual temperature increase, changes in surface water along the herder migration routes, and the capacity of the herders to sustain the levels of mobility needed for accessing viable pasture grounds for the reindeer while attending to their own health care needs. We further argue that this temporal shift is related to the warming temperatures on Kanin. However, similar direction of hydrologic changes has been reported in other regions of the circumpolar north that also have warmed to a comparably high level during the same period ([@CIT0002],[@CIT0017]) and have negatively affected access to health care across the Arctic.
In conclusion, this investigation demonstrated the feasibility of future studies that rely on the participation and the local knowledge of the indigenous community and of local medical practitioners and on remotely sensed data. Given that researchers and local residents are observing similar changes in other parts of the circumpolar north, the approach demonstrated here is highly promising if applied to other Arctic contexts (for example, in connection with the coastal erosion and thermokarst processes and their implications for subsistence-based livelihoods and assess to health care).
Disparities in access to health care have been shown to negatively affect health outcomes elsewhere ([@CIT0017]--[@CIT0020]), and we suggest that new health care delivery models, responsive to the new reality of rapid Arctic environmental change, are necessary to improve the health of indigenous circumpolar peoples. One research direction with a potential to improve access to health care and to reduce the risk of injuries is to provide near-real-time new water detection maps directly to the subsistence-based circumpolar communities via an Internet portal. This approach may prove effective in visualizing environmental change and in planning future health services provision in the rapidly changing Arctic environment.
We would like to thank the Nenets community on Kanin and the local health practitioners that serve it. Without their generous help, this study would not be possible.
Conflict of interest and funding {#S0008}
================================
Part of this research was funded by a grant/cooperative agreement from the National Oceanic and Atmospheric Administration (NOAA), NA10OAR4310210. The views expressed herein are those of the authors and do not necessarily reflect the views of NOAA or any of its sub-agencies. This study was approved by the institutional review board of the State University of New York Downstate.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#S1}
============
The importance of the biological properties of antibodies to specifically engage a target of choice and activate complement and Fc gamma receptors (FcγR) on immune cells ([@B1]) is currently more and more recognized in modern medicine. For cancer therapies using tumor targeting antibodies, strong effector functions are preferred ([@B2]). Various strategies have been exploited to generate antibodies that are more effective than wild-type human IgG1 isotype ([@B3]). These include fusions with toxic molecules and incorporations of mutations that enhance affinities to FcγR. Possible drawback of such modifications is the introduction of foreign immunogenic epitopes that can result in anti-drug antibodies that may neutralize the drug. This can be circumvented by using non-immunogenic natural variations, found in all individuals. The prototypic variation of this kind are glyco-engineered IgG1 antibodies without fucose with elevated FcγRIIIa affinities ([@B4], [@B5]), which have already found its way to therapeutic antibodies on the market ([@B6]).
This fucose residue is part of a conserved glycan on asparagine 297 in the Fc domain of immunoglobulin G (IgG). This glycan is important for the quaternary structure of the Fc part, since its removal abrogates binding of FcγR and C1q and hence the antibody's effector functions ([@B7]--[@B9]). In addition to affecting the Fc structure and thereby recognition by these effector molecules, the Fc-glycan also affects binding to FcγRIIIa and FcγRIIIb through a glycan--glycan interaction ([@B10], [@B11]). This is because of a unique glycan found in human FcγRIIIa and FcγRIIIb at position 162 that interacts directly with the Fc-glycan within the IgG-Fc cavity ([@B11]).
The N297 glycan is a bi-antennary complex glycan composed of a constant part with a core consisting of *N*-acetylglycosamines and mannoses and can be found in human serum with variable levels of core fucose, bisecting *N*-acetylglycosamine, galactose, and terminal sialic acids ([@B12]). The N-glycans of total serum/plasma IgG consists on average of high fucose levels (95%), low bisection (15%), intermediate levels of galactose (45%), and low sialic acid (10%) ([@B12]). The variable assembly of the glycans amounts to at least 20 different glycoforms (a term used here to describe one unique glycan combination) for each IgG subclass being found in serum, with \~8 of them accounting for 90% of the total abundance ([@B12]). The composition of total IgG glycosylation can change upon certain settings, where galactosylation and sialylation increase with pregnancy ([@B12], [@B13]). Changes in total IgG are also observed in various clinical settings, with a low level of galactosylation and sialylation associated both with increasing age and autoimmune diseases ([@B12]--[@B15]).
We and others have shown that IgG-Fc glycosylation changes of antigen-specific IgG can occur that correlate with disease outcome ([@B16]--[@B20]). This includes both auto- and alloimmune disorders, including fetal neonatal immune thrombocytopenia (FNAIT), immune thrombocytopenia, and hemolytic disease of the fetus and newborn (HDFN) ([@B16]--[@B18], [@B21], [@B22]). In particular, we have found that immune responses against red blood cell (RBC) and platelets, either transfused or during pregnancy, can be characterized with extremely low fucose (down to 10%), high galactose (up to 80%), and elevated sialylation levels (\~35%). Notably, lowered Fc-fucosylation ([@B17], [@B21]), but also elevated Fc-galactosylation ([@B18]), seemed to correlate with elevated blood cell destruction, severity of anemia or bleeding for RBCs and platelets, respectively. Whereas the increased pathogenicity associated with lowered fucosylation could be explained by the resulting elevated FcγRIIIa and/or FcγRIIIb activity ([@B17], [@B23]), the functional reasons---if any---behind the association with elevated galactosylation remained enigmatic.
The effect of Fc-bisection and -sialylation on human FcγR binding, if any, has been studied in even less detail, although binding to the human FcγRIIIa does not seem to be affected by sialylation ([@B24]). Whether these glycan changes influence binding to C1q, and subsequent complement activation, has not been studied in detail ([@B25], [@B26]). A drawback of all these studies is that the impact of the glycan changes was studied changing only individual end groups, without investigating the possibility that the context of the other glycan changes may have an effect on the antibody effector functions.
The complexity of the glycan-assembly makes investigation into their biological relevance extremely difficult. Previous attempts have generated a handful of defined glycoforms and tested binding to part of the FcγR-repertoire, but a systematic analysis for all possible glycan changes and effector mediators, FcγRs and complement, has never been achieved ([@B24], [@B26]--[@B30]). This information could provide the insight in working mechanisms of IgG-based treatments and allow meaningful clinical evaluation of the activity of potentially pathological antibodies such as in FNAIT and HDFN. We have, therefore, developed a set of glyco-engineering tools which specifically alter one of the N-glycan end groups ([@B31]) and in the present study we combined these tool to create 20 different natural glycoforms to systematically investigate them with regard to FcγR binding, antibody-dependent cellular cytotoxicity (ADCC), complement binding, and activation.
Materials and Methods {#S2}
=====================
Human Samples {#S2-1}
-------------
Peripheral blood from anonymous, healthy volunteers was obtained with informed, written consent in accordance with Dutch regulations. This study was approved by the Sanquin Ethical Advisory Board in accordance with the Declaration of Helsinki.
Heparinized blood samples were used for isolation of peripheral blood mononuclear cells (PBMCs) or RBCs. NK cell isolation was only performed with blood from well-genotyped donors who do not express FcγRIIc ([@B32]) to exclude any possible effects of this receptor. Serum was obtained by allowing blood without anticoagulants to coagulate for 1 h at room temperature (RT) and collecting the supernatant after centrifugation at 950 × *g* for 10 min. Serum of three different volunteers was combined to create a serum pool.
Strains and Reagents {#S2-2}
--------------------
*Escherichia coli* strain DH5α was used for recombinant DNA work. Restriction endonucleases, DNA modification enzymes were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Oligonucleotides were obtained from Geneart (Thermo Fisher Scientific) or Integrated DNA Technologies (Coralville, IA, USA).
IgG1 Expression Vector Constructs {#S2-3}
---------------------------------
Variable (V) genes for anti-human RhD (anti-D clone 19A10) heavy and light chain were sequenced from a single human B cell from a hyper immunized donor ([@B33]). A single-gene vector containing anti-D or anti-TNP IgG1 heavy- and kappa light-chain-encoding sequences were cloned as described previously by Kruijsen et al. ([@B34]) into a pEE14.4 (Lonza, Basel, Switzerland) expression vector. For both anti-TNP and anti-D IgG, a single expression vector was generated. In brief, the codon-optimized V gene for both heavy and light chain, including 5′-HindIII and 3′-NheI or 5′-HindIII and 3′-XhoI restriction sites respectively, Kozak sequence, and HAVT20-leader sequence, were designed and ordered from Geneart (Thermo Fisher Scientific). The HindIII-NheI or HindIII-XhoI fragments for the codon-optimized heavy or light chain were ligated into γ or κ constant region flanking 3′-EcoRI restriction site, respectively. The HindIII--EcoRI fragment for the codon-optimized light chain was ligated into pEE14.4 (Lonza), and the HindIII--EcoRI fragment for the heavy chain was ligated into pEE6.4 (Lonza). A single-gene vector encoding IgG1 was subsequently generated by ligation of the BamHI--NotI fragment from pEE6.4 (including a cytomegalovirus promoter), IgG1 heavy chain, and poly (A) into the light-chain-encoding pEE14.4 vector.
IgG1 Production and Glyco-Engineering {#S2-4}
-------------------------------------
IgG1 production in human embryonic kidney (HEK) F cells and purification using protein A affinity chromatography was performed as described previously by Kruijssen et al. ([@B34]) Glyco-engineering of IgG1 was optimized as described by Dekkers et al. ([@B31]) In short, to decrease either fucosylation or galactosylation, 0.4 mM 2-deoxy-2-fluoro-[l]{.smallcaps}-fucose (2FF) (Carbosynth, Berkshire, United Kingdom) or 1 mM 2-deoxy-2-fluoro-[d]{.smallcaps}-galactose (2FG) (Carbosynth), respectively, was added to the cell suspension 4 h post transfection. To increase bisecting GlcNAc, 1% pEE6.4 + GNTIII encoding mannosyl (beta-1,4-)-glycoprotein beta-1,4-*N*-acetylglucosaminyltransferase (GNTIII) enzyme was co-transfected with 99% IgG1-κ HC + LC vector. To increase galactose, 1% pEE6.4 + B4GALT1 encoding β-1,4-galactosyltransferase 1 (B4GALT1) enzyme was co-transfected with 99% IgG1 vector and 5 mM [d]{.smallcaps}-galactose (Sigma Aldrich, Saint Louis, MO, USA) was added to the cell suspension 1 h before transfection. To increase sialylation, the level of galactosylation must also be elevated as sialic acid is the terminal sugar group with galactose residues as substrate. Thus, 1% pEE6.4 + B4GALT1 and 2.5% pEE14.4 + STGALT encoding β-galactoside alpha-2,6-sialyltransferase 1 (ST6GALT) were both co-transfected 96.5% IgG1 vector and 5 mM [d]{.smallcaps}-galactose was added to the cell suspension 1 h before transfection. To further increase sialylation, *in vitro* sialylation (ivs) was performed on the purified *in vivo* sialylated IgG created using the previous method. Recombinant human α-2,6-sialyltransferase (Roche, Basel, Switzerland) and cytidine-5′-monophospho-*N*-acetylneuraminic acid (CMP-NANA) (Roche) were incubated at 37°C for 24 h with purified IgG1 with already *in vivo* enhanced galactose and sialic acid (as described above), after incubation samples were re-purified with protein A, as described previously ([@B31], [@B34]).
Mass Spectrometry Analysis {#S2-5}
--------------------------
Immunoglobulin G-Fc glycan composition of produced IgG1 was determined by mass spectrometry as described previously by Dekkers et al. ([@B31]) Trypsin-digested glycopeptide samples were analyzed by nanoLC--ESI--QTOF--MS. The separation was performed on an RSLCnano Ultimate 3000 system (Thermofisher, Breda, the Netherlands) with a gradient pump, loading pump and an autosampler. 250 nl of sample was injected and washed on a Dionex Acclaim PepMap100 C18 trap column (5 mm × 300 µm i.d.; Thermofisher) for 1 min with 0.1% TFA at a flow rate of 25 µl/min. The sample was then separated on an Ascentis Express C18 nanoLC analytic column (50 mm × 75 µm i.d.; 2.7-µm fused core particles; Supelco, Bellefonte, PA) with a flow rate of 0.9 µl/min using linear gradient as described in Ref ([@B30]). The resulting co-elution of the different glycoforms of the IgG1-Fc glycosylation site warrants fair comparison by ensuring identical ionization conditions for the various glycopeptide species. The LC was coupled to the MS detector *via* a CaptiveSpray source with a NanoBooster (Bruker Daltonics, Bremen, Germany). The latter enriched the N~2~ flow (3 l/min) with CH~3~CN (pressure 0.2 bar), resulting in increased sensitivity. The samples were ionized in positive ion mode at 1,100 V. The Maxis Impact quadrupole-TOF--MS (micrOTOF-Q, Bruker Daltonics) was used as detector. MS1 spectra were collected at a frequency of 1 Hz with a scan range of *m/z* 550--1,800. The mass spectrometric data were calibrated internally in DataAnalysis 4.0 (Bruker Daltonics) using a list of known IgG glycopeptide masses. MSConvert (Proteowizard 3.0) ([@B35]) was used to convert the data files to mzXML format, and an in-house alignment tool ([@B36]) was used to align the retention times of the data files. The highest intensity of selected peaks (within an *m/z* window of ±0.2 and within a time window of ±15 s surrounding the retention time) was extracted using the in-house developed 3D Max Xtractor software tool. If above a signal:background ratio of 3, the background-subtracted area of the first three isotopic peaks of each glycopeptide in both 2+, 3+, and 4+ charge state were summed, and this summed value was then divided by the total summed value of all IgG1 glycopeptides to arrive at a percentage for each glycopeptide. From these percentages, we calculated several derived traits using the following formulas: fucosylation (H3N3F1 + H4N3F1 + H5N3F1 + H6N3F1 + G0F + G1F + G2F + H6N4F1 + G0FN + G1FN + G2FN + H6N5F1 + H4N3F1S1 + H5N3F1S1 + H6N3F1S1 + G1FS + G2FS + H6N4F1S1 + G2FS2 + G1FNS + G2FNS + H6N5F1S1 + G2FNS2), bisection (H6N4F1 + G0FN + G1FN + G2FN + H6N5F1 + H6N4F1S1 + G1FNS + G2FNS + H6N5F1S1 + G2FNS2 + H6N4 + G0N + G1N + G2N + H6N5 + H6N4S1 + G1NS + G2NS + H6N5S1 + G2NS2), galactosylation \[(H4N3F1 + H5N3F1 + G1F + H6N4F1 + G1FN + H6N5F1 + H4N3F1S1 + H5N3F1S1 + H6N3F1S1 + G1FS + H6N4F1S1 + G1FNS + H6N5F1S1 + H4N3 + H5N3 + H6N3 + G1 + H6N4 + G1N + H6N5 + H4N3S1 + H5N3S1 + H6N3S1 + G1S + H6N4S1 + G1NS + H6N5S1) \* 0.5 + G2F + G2FN + G2FS + G2FS2 + G2FNS + G2FNS2 + G2 + G2N + G2S + G2S2 + G2NS + G2NS2\], sialylation \[(H4N3F1S1 + H5N3F1S1 + H6N3F1S1 + G1FS + G2FS + H6N4F1S1 + G1FNS + G2FNS + H6N5F1S1 + H4N3S1 + H5N3S1 + H6N3S1 + G1S + G2S + H6N4S1 + G1NS + G2NS + H6N5S1) \* 0.5 + G2FS2 + G2FNS2 + G2S2 + G2NS2\], hybrid-types (H5N3F1 + H6N3F1 + H6N4F1 + H6N5F1 + H5N3F1S1 + H6N3F1S1 + H6N4F1S1 + H6N5F1S1 + H5N3 + H6N3 + H6N4 + H6N5 + H5N3S1 + H6N3S1 + H6N4S1 + H6N5S1), and high-mannose (H5N2 + H6N2 + H7N2 + H8N2 + H9N2). For some of the minor hybrid-type glycans, it could not be determined conclusively whether a galactose or a bisecting *N*-acetylglucosamine was present, so an educated guess was made based on structural knowledge (for instance, since the hybrid glycan H6N4F1 is elevated in GNTIII-co-transfected HEK cell-derived IgG samples, it is likely to be a bisected species rather than triantennary).
High-Performance Liquid Chromatography (HPLC) {#S2-6}
---------------------------------------------
Protein A purified IgG was analyzed for monomeric and dimeric IgG on a Superdex 200 10/300 gel filtration column (30 cm, 24 ml, 17-15175-01, GE Healthcare, Little Chalfont, United Kingdom) connected to an Äkta explorer (GE Healthcare) HPLC system at RT with a flow rate of 0.5 ml/min and PBS as running buffer. Elution profiles were obtained by recording the absorbance at 215 nm.
Human FcγR Constructs {#S2-7}
---------------------
Human FcγR constructs \[FcγRIa (HIS tag), FcγRIIa (131His, Biotinylated, and 131Arg, Biotinylated), FcγRIIb (Biotinylated), FcγRIIIa (158Phe, Biotinylated, and 158Val, Biotinylated) and FcγRIIIb (NA2, HIS tag)\] for surface plasmon resonance (SPR) analysis were obtained from Sino biological (Beijing, China). To further include all human FcγRs, a fusion Fc--FcγR construct composed of the extracellular domain of the FcγRIIIb in both allotypes followed by a Fc domain was created. To create the fusion Fc--FcγRIIIb constructs the amino acid code of the extracellular domain of either FcγRIIIb of NA1 allotype or FcγRIIIb NA2 allotype ([@B37]) (NCBI reference sequence NP_000561.3), and IgG2 Fc domain, composed of a human IgA1a hinge, human IgG2 Fc CH2 and CH3 domains including mutations deleting the Fc-glycan (N297A) and introducing a C-terminal biotinylation tag (BirA) were reverse translated and codon optimized at Geneart. DNA was ordered (Integrated DNA technologies, Coralville, IA, USA) and cloned into pcDNA3.1 (Invitrogen, Carlsbad, CA, USA) expression vector using flanking HindIII and EcoRV restriction sites. A model of the construct and sequences are displayed in Figures S7A,B in Supplementary Material. The construct was produced and purified as described previously ([@B31]). After purification the protein was site-specifically biotinylated on the BirA tag using BirA enzyme as described by Rodenko et al. ([@B38]). For biotinylation of 1 µM FcγR protein 0.00657 µM BirA ligase was used. After biotinylation overnight at 25°C, the FcγR sample was buffer-exchanged and subsequently concentrated in PBS pH 7.4 using Amicon Ultra centrifugal filter units (MWCO 30 kDa) (Merck, Millipore, Darmstadt, Germany). The quality of the Fc-Fusion receptors was confirmed by comparing the binding of normally glycosylated IgG1 to the acquired his-tagged receptor (Sino-biological) and in-house made Fc-Fusion of the same allotype (NA2) (Figures S7C,D in Supplementary Material).
Surface Plasmon Resonance {#S2-8}
-------------------------
Surface plasmon resonance measurement were performed as described by Dekkers et al. ([@B39]). All biotinylated FcγR were spotted using a Continuous Flow Microspotter (Wasatch Microfluidics, Salt Lake City, UT, USA) onto a single SensEye G-streptavidin sensor (Ssens, Enschede, Netherlands) allowing for binding affinity measurements of each antibody to all FcγR simultaneously on the IBIS MX96 (IBIS Technologies, Enschede, Netherlands) as described by de Lau et al. ([@B40]). The biotinylated FcγRs were spotted in threefold dilutions, ranging from 100 to 3 nM for FcγRIIb and fusion FcγRIIIb-IgG2-Fc. All the other FcγRs were spotted in threefold dilutions, ranging from 30 to 1 nM in PBS 0.0075% Tween-80 (Amresco), pH 7.4. The IgGs were then injected over the IBIS at 1.5 dilution series starting at 5.9 nM until 506.25 nM or 0.9 nM until 2,000 nM, when necessary, in PBS in 0.075% Tween-80. For FcγRI affinity and FcγRIIIb control measurements, his-tagged FcγRI or FcγRIIIb was used. Biotinylated anti-His-tagged antibody (Genscript Piscataway, NJ, USA) was spotted in threefold dilutions, ranging from 30 to 1 nM. Before every IgG injection, 50 nM his-tagged FcγR was injected. The IgGs were then injected over the IBIS at threefold dilution series starting at 0.41 nM until 100 nM for FcγRI and 94 nM until 3,000 nM for FcγRIIIb. Regeneration after every sample was carried out with acid buffer (10 mM Gly--HCl, pH 2.4). Calculation of the dissociation constant (*K*~D~) was done using an equilibrium analysis by linear intrapolation to Rmax = 500 ([@B41]). Analysis and calculation of all binding data were carried out with Scrubber software version 2 (Biologic Software, Campbell, ACT, Australia) and Microsoft Office Excel 2013.
NK Cell-Mediated ADCC {#S2-9}
---------------------
NK cells were isolated from Ficoll-Plaque™-Plus (GE Healthcare) gradient obtained PBMCs by a CD56 magnetic-activated cell separation isolation kit (Miltenyi Biotec, Leiden, The Netherlands), according to manufacturer's description. D + RBCs were isolated and labeled with radioactive chromium (100 μCi ^51^Cr, PerkinElmer, Waltham, MA, USA) at 10^9^ cells/ml. An amount of 10^5^ erythrocytes were incubated with NK cells for 2 h at 37°C in a 2:1 ratio in Iscove's modified dulbecco's medium (IMDM, Gibco, Thermo Fisher Scientific) supplemented with 10% fetal calf serum (FCS, Bodinco, Alkmaar, The Netherlands) and anti-D IgG1-glycoforms at a total volume of 100 µl. To determine 100% lysis, 2.5% saponine (Fluka, Sigma Aldrich) was added to RBC in control wells and spontaneous lysis (sp) was determined by incubation of RBC without NK cells. Supernatants were collected and released ^51^Cr was quantified in a Packard Cobra II Auto-Gamma Counter Model D5005 (PerkinElmer). Percentage cytotoxicity was determined by the following formula: $\text{ADCC}\left( \% \right) = \frac{\text{counts sample} - \text{counts sp}}{\text{counts 100\%} - \text{counts sp}} \times 100$, each value consisted of at least three individual sample wells.
Complement Deposition ELISA {#S2-10}
---------------------------
A 2.4-mM 2,4,6-trinitrobenzenesulfonic acid (TNBS) (Sigma-Aldrich) solution was added to 20 mg human serum albumin (HSA) diluted to 20 mg/ml (Sanquin, Amsterdam, The Netherlands) in 0.2 M NA~2~HPO~4~ (Merck, Millipore) and incubated 30 min at RT. To remove unbound TNBS, the solution was dialyzed (1:2,000) using a dialysis cassette (Thermo Fisher Scientific Slide-A-Lyzer G2 cassette, 10K MWCO) for 1.5 h at RT against PBS and additionally overnight at 4°C to obtain HSA-TNP.
To coat, maxisorp plates (Thermo Scientific, Nunc flat-bottom 96-well plate) were incubated o/n at RT with 20 µg/ml HSA-TNP in PBS. The plates were washed 5× with PBS + 0.1% tween-20 (Sigma-Aldrich) (wash buffer) using an ELISA washer (Biotek, 405 LSRS). All following washing steps were done similarly. The IgG samples were diluted in 100 µl PBS/plx \[PBS + 0.1% poloxamer (Sigma-Aldrich, poloxamer 407)\] per well and incubated for 1.5 h at RT. The plates were washed and incubated with 100 µl 1:35 serum pool in VB^+/+^/plx {veronalbuffer \[3 mM Barbital (Sigma-Aldrich), 1.8 mM Sodium-Barbital (Sigma Aldrich), 0.146 M NaCl (Fagron, Capelle aan den Ijssel, The Netherlands), pH 7.4\] + 10 mM CaCl~2~ (Merck) + 2 mM MgCl~2~ (Merck) + 0.1% poloxamer} for 1 h at RT. When C1q was blocked, 10 min prior to addition of serum to the ELISA plate, anti-C1q-85 blocking antibody ([@B42]) was added to the VB^+/+^/plx + 1:35 serum solution in a 1:2 molar ratio of C1q:anti-C1q-85 with final concentration of 8.57 µg/ml anti-C1q-85. The plates were washed and 100 µl with either 2 µg/ml biotinylated anti-C1q-2 ([@B42]), 0.5 µg/ml biotinylated anti-C4-10 ([@B43]), 0.6 µg/ml biotinylated anti-C3-19 ([@B44]), or 1 µg/ml HRP labeled anti-human IgG (Sanquin, Peliclass) in PBS/plx was added to respectively detect C1q, C4b, C3b, or IgG deposition and incubated for 1 h at RT. The plates were washed, C1q, C4b, and C3 plates were incubated with 100 µl 0.2 µg/ml strep-poly HRP (Sanquin, Peliclass) (C1q) or 0.25 µg/ml strep-HRP (Sigma-Aldrich) (C4b and C3b) in PBS/plx for 1 h at RT. The plates were washed and developed for 5--10 min using 100 µl TMB mix composed of 0.11 M NaAc (pH 5.5) (Merck), 0.1 mg/ml 3,3′,5,5′-Tetramethylbenzidine (Merck) and 0.003% H~2~O~2~ (Merck) and the reaction was stopped with the addition of 100 µl 2 M H~2~SO~4~ (Merck). The optical density (OD) was measured at A450 nm using a plate reader (Biotek, Synergy 2, Winooski, VT, USA).
The results were analyzed with a parallel line assay in Microsoft Office Excel ([@B45]). We assessed the potency of the glycoforms relative to a standard, an independently titrated unmodified IgG1; these values were expressed as percentages relative to the unmodified glycoform.
Complement-Mediated Lysis {#S2-11}
-------------------------
Fifty microliters of washed, packed, D^+^ RBCs obtained from heparinized blood were mixed with 350 µl 0.313 mM TNBS in 0.15 M Na~2~HPO~4~, pH 8.8 and incubated for 10 min at RT. TNPylated RBCs were centrifuged for 2 min at 350 × *g* and washed two times with PBS. RBC were resuspended into VBG^+/+^ \[VB^+/+^ + 0.05% w/v gelatin (Sigma-Aldrich)\]. Anti-TNP IgG1 was serially diluted in VBG^−/−^ (3 mM Barbital, 1.8 mM Sodium-Barbital, 0.146 M NaCl, pH 7.4, 0.05% w/v gelatin). In round bottom plates to a final volume of 100 µl we combined the diluted IgG1, 10% serum, \~4.5 × 10^6^ RBC, and a glass bead (2 mm, Merck) to ensure mixing of the solution during incubation (1:1 final ratio VBG^−/−^:VBG^+/+^). This amount of RBC was taken to ensure the 100% absorbance between 1.8 and 2.2 delta (Δ) A412--A690 nm. The plates were incubated for 90 min at 37°C while shaking at 150 rpm (Orbital incubator S150, 16 mm shaking diameter). After incubation, 1.25% w/v saponine was supplemented to the 100% control wells, 100 µl VBG^−/−^ was added to all wells and the plates were centrifuged for 2 min at 350 × *g*. Subsequently, 150 µl of supernatant was transferred into a separate plate and OD was measured at Δ A412--A690 nm using a plate reader. The percentage of lysed cells was calculated as follows: $\text{Lysis}\left( \% \right) = \frac{\text{OD sample-OD spontaneous}}{\text{OD 100-OD spontaneous}} \times 100$. In GraphPad Prism, we calculated the half maximal effective concentration (EC~50~) for each replicate of the different glycoforms using a non-linear fit for normalized response with a variable slope and combined these to an average EC~50~.
Statistical Analysis {#S2-12}
--------------------
Statistical analyses were performed using GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla, CA, USA). The level of significance was set at *p* \< 0.05 using two-tailed tests.
Results {#S3}
=======
Recapitulation of All 20 Major Different Glycoforms Found in Human Plasma {#S3-1}
-------------------------------------------------------------------------
Human IgG1, produced in HEK cells, shows complex-type bi-antennary glycans similar to IgG from normal human plasma (Figure [1](#F1){ref-type="fig"}A) ([@B31], [@B46]). More specifically, without any modification ("Unmodified," box labeled "U" in the *x*-axis legend, Figures [1](#F1){ref-type="fig"}B--E) HEK-derived IgG1 N-glycans feature high fucosylation, low bisection, intermediate-level galactosylation, and low sialylation (Figures [1](#F1){ref-type="fig"}B--E). We previously developed six glyco-engineering tools which can be implemented upon protein production, as we recently described ([@B31]). These were aimed to decrease fucosylation, increase bisection, decrease or increase galactosylation, or increase sialylation. In the present study, these tools were combined and used in all possible combinations during the transient transfection in HEK cells, which resulted in the anticipated glycosylation changes and allowed us to produce the 20 major glycoforms present in human serum. Only minor unanticipated effects (Figures [1](#F1){ref-type="fig"}B--E), were observed. A slight increase in galactosylation upon overexpression of beta 1,4-*N*-acetylglucosaminyltransferase III (GntIII) to increase bisection (e.g., 28 to 36% upon GntIII expression)---but this was only observed in samples with low starting-levels of galactosylation (Figures [1](#F1){ref-type="fig"}C,D). Some of the tools caused a minor increase (\<21%) in high-mannose or hybrid glycan species (Figure S1 and Table S1 in Supplementary Material) ([@B31]). Using the glyco-engineering tools the most extreme levels were reached for fucose and galactose (Figures [1](#F1){ref-type="fig"}B,D), bisection was increased up to 60%, and sialylation never reached over half of what was possible by the underlying galactose (\~40%). The level of sialylation was, therefore, further increased using *in vitro* sialylation as described before (up to \~70%) (Figure [1](#F1){ref-type="fig"}E; Table [1](#T1){ref-type="table"}) ([@B31], [@B47], [@B48]). All in all, this resulted in 20 combinations and markedly different glycoforms. All 20 glycoforms were produced as two panels of IgG1, specific for the RhD (anti-D) antigen or 2,4,6-trinitrophenyl hapten (anti-TNP) ([@B34]), with both panels showing highly comparable glycosylation patterns depending of the glyco-engineering tools applied (Table [1](#T1){ref-type="table"}; Table S1 in Supplementary Material). To avoid any possible confounding effects of Fab glycosylation on IgG function, we used anti-D and anti-TNP with variable domains sequences devoid of N-linked glycosylation sites.
![Recapitulation of 20 different immunoglobulin G (IgG) glycoforms by glyco-engineering. **(A)** Model of IgG with glycan at position N297 in the Fc domain and composition of the glycan. **(B--E)** Degree of derived glycan traits as reached by the different glyco-engineering tools: 2FF, 0.4 mM 2-deoxy-fluoro-[l]{.smallcaps}-fucose; GntIII, 1% GntIII co-transfection; 2FG, 1 mM 2-deoxy-fluoro-[d]{.smallcaps}-galactose; B4galT1/Dgal, 1% B4GALT1 co-transfection and 5 mM [d]{.smallcaps}-galactose; ST6GALT, 2.5% ST6GALT co-transfection, *in vitro* sial, treatment of IgG with recombinant ST6GALT and CMP-NANA substrate. The data represent the mean and SEM of at least two combined independent experiments; \*, \*\*, \*\*\*, and \*\*\*\* denote a statistical significance of *p* ≤ 0.05, *p* ≤ 0.01, *p* ≤ 0.001, and *p* ≤ 0.0001, respectively, as tested by one-way ANOVA against unmodified IgG1, using Dunnett's multiple comparisons test. U: unmodified glycoform.](fimmu-08-00877-g001){#F1}
######
Comprehensive list of glycopeptide degrees of complex glycans found in the glyco-engineered IgG1 batches of anti-D and anti-TNP specificity.
------------------------------
------------------------------
*2FF, 0.4 mM 2-deoxy-2-fluoro-[l]{.smallcaps}-fucose; 2FG, 1 mM 2-deoxy-2-fluoro-[d]{.smallcaps}-galactose; GNT3, co-transfection of 1% GNT3 vector; B4GALT1/[d]{.smallcaps}-galactose, co-transfection of 1% B4GALT1 vector and addition of 5 mM [d]{.smallcaps}-galactose; ST6GALT, co-transfection of 2.5% ST6GALT vector; *in vitro* sialylation, treatment of sample with recombinant ST6GALT and CMP-NANA; G, number of galactoses; F, presence of a core fucose; N, presence of a bisecting N-acetylglucosamine; S, number of *N*-acetylneuraminic (sialic) acids*.
*These resulted in significantly different derived glycosylation traits (fucosylation, bisection, galactosylation, sialylation, high-mannose, hybrid-type), which are calculated from the relative abundances of individual *N*-glycans*.
*The shading of cells indicates, for each glycoform the lowest to highest abundance of glycopeptides, respectively, from light to dark*.
*^a^For complex immunoglobulin G glycans, we use a nomenclature which assumes structural knowledge of the glycans based on literature*.
Binding of IgG Glycome to Human FcγR {#S3-2}
------------------------------------
We next used the IBIS MX96 biosensor system, as described in Dekkers et al. ([@B39]), capable of analyzing the binding of up to 48 different receptor ligand interactions in parallel by SPR, to probe the affinity of all IgG1-glycoforms to all human FcγRs and their allotypes affecting IgG binding (Table S2 in Supplementary Material) ([@B49]). The antibodies used for these experiments (anti-D) showed no signs of dimers or multimers (Figure S2 in Supplementary Material). The binding affinities of unmodified IgG1 to the different receptors resembled those reported earlier (Table [2](#T2){ref-type="table"}) ([@B49]). We considered significant changes in the apparent *K*~D~ of more than twofold from unmodified IgG to be potentially meaningful changes and within the scope of the SPR method, using a simplified 1:1 Langmuir model that does not fully represent the actual interaction which is more complicated ([@B39]). No significant effects of glycan changes above twofold were seen on the binding to FcγRIa, FcγRIIa (neither the H131- or the R131-allotype), or FcγRIIb/c (Figures [2](#F2){ref-type="fig"}A--D). However, marked changes were seen for all FcγRIII-isoforms. Reduction of fucose resulted in enhanced binding to all FcγRIII species by approximately 10- to 20-fold depending on the type- and allotype (Figures [2](#F2){ref-type="fig"}E--H), as reported ([@B4], [@B5]). Importantly, addition of galactose consistently enhanced binding of hypo-fucosylated IgG1 for all FcγRIIIa allotypes, doubling the effect of hypo-fucosylation alone (Figures [2](#F2){ref-type="fig"}E,F). This effect was also seen for allotypes of FcγRIIIb, but less strong and only for IgG1 that was bisected in addition to hypo-fucosylated (Figures [2](#F2){ref-type="fig"}G,H). Further sialylation of low-fucosylated galactosylated IgG1 had little additional effect on the binding to FcγRIII, except for hypo-fucosylated and bisected IgG1 for both allotypes of FcγRIIIa and FcγRIIIb NA2, where sialylation cause a significant decrease in binding. Taken together, glycan changes in the IgG-Fc only affect binding to FcγRIIIa and FcγRIIIb, with a major effect of hypo-fucosylation increasing binding to FcγRIIIa/b that was boosted by galactosylation. Bisection only appeared to indirectly affect binding when occurring in conjunction with sialylation, causing a slight decreased binding to FcγRIIIa/b to otherwise hypo-fucosylated and galactosylated IgG.
######
Affinity of unmodified IgG1 (U) to the different FcγRs.
Mean affinity SEM
---------------- --------------- ----------------
FcγRI 3.0 × 10^−9^ ±8.7 × 10^−10^
FcγRIIa^131H^ 3.8 × 10^−7^ ±1.3 × 10^−8^
FcγRIIa^131R^ 4.8 × 10^−7^ ±2.8 × 10^−8^
FcγRIIb 2.7 × 10^−6^ ±1.1 × 10^−7^
FcγRIIIa^158F^ 1.3 × 10^−6^ ±9.5 × 10^−8^
FcγRIIIa^158V^ 2.4 × 10^−7^ ±1.0 × 10^−8^
FcγRIIIb NA1 3.2 × 10^−6^ ±4.7 × 10^−7^
FcγRIIIb NA2 2.8 × 10^−6^ ±1.0 × 10^−7^
*Affinity in K~D~, as measured by SPR*.
{#F2}
FcγRIIIa-Mediated ADCC Is Steered by Fucosylation and Galactosylation {#S3-3}
---------------------------------------------------------------------
We next tested the efficacy of these anti-D IgG1 antibodies to mediate ADCC against RBC. Curiously, no NK cell-mediated induction of ADCC was seen with any fucosylated IgG1 at any concentration tested (Figures [3](#F3){ref-type="fig"}A,B; Figure S3 in Supplementary Material). Only hypo-fucosylated IgG1 induced ADCC in variable degrees depending on the glycosylation (Figures [3](#F3){ref-type="fig"}A,B). The observed level of ADCC were in line with the binding results obtained by SPR for each of the FcγRIIIa allotypes (Figure [3](#F3){ref-type="fig"}C), confirming the essential role of both hypo-fucosylation and elevated galactosylation for increased FcγRIIIa-binding and effector functions. Again, sialic acid had a minor but significant negative effect, especially for the bisected, hypo-fucosylated, and galactosylated IgG1 (Figures [3](#F3){ref-type="fig"}A,B). Remarkably, the well-known allotypic differences in affinity were confirmed by our SPR experiments, but not by the functional NK cell-mediated ADCC.
{#F3}
Galactosylation and Sialylation Direct Complement Binding and Activation {#S3-4}
------------------------------------------------------------------------
We then tested the effect of IgG-Fc glycosylation on C1q binding and subsequent complement activation, using the anti-TNP panel of IgG1 antibodies as anti-D does not fix complement. The efficiency of C1q binding to TNP-lated human serum albumin (TNP-HSA) and subsequent C4b deposition was titrated by serial dilution (Figure S4 in Supplementary Material). All glycovariants of anti-TNP bound TNP-HSA equally well (Figure S4A in Supplementary Material), but C1q binding and C4b deposition differed profoundly for the different glycoforms (Figure S4B in Supplementary Material). The relative C1q binding and C4b deposition were then calculated (Figures [4](#F4){ref-type="fig"}A,B, respectively). Both data sets suggested that elevated galactosylation and sialylation positively influenced complement activity. This activity was fully depended on the classical pathway with no influence of the mannan-binding lectin- or the alternative pathway, as C4b and C3b deposition, were completely blocked by an anti-C1q blocking antibody (Figure S5 in Supplementary Material). We then determined if this also translates into more efficient complement-dependent cytotoxicity (CDC) by analyzing complement-dependent lysis of TNP-labeled RBC (Figure [4](#F4){ref-type="fig"}C; Figure S6 in Supplementary Material). The level of C1q binding of each glycoform correlated well with the C4b deposition (Figure [5](#F5){ref-type="fig"}A) and with the obtained EC~50~ of CDC (Figure [5](#F5){ref-type="fig"}B). The level of galactosylation of each glycoform also showed a direct relationship with the efficacy of C1q binding, and EC~50~ (Figures [5](#F5){ref-type="fig"}C,D). In conclusion, the degree of galactosylation, but also sialylation of the IgG1-Fc N-glycan directly steers the antibody's efficacy to stimulate complement deposition and CDC.
{#F4}
{#F5}
Discussion {#S4}
==========
We have previously created an orthogonal set of glyco-engineering tools ([@B31]) which we now combined to create 20 glycovariants of human IgG1, representing natural variants found in human plasma IgG, including extreme glycoforms found for examples in patients with FNAIT and HDFN ([@B17], [@B18], [@B21], [@B22]). These variants were investigated for their functional capacity to engage and activate FcγR and complement.
Of the FcγRs, we only observed an effect of glycosylation on binding to the FcγRIII-family of receptors, both FcγRIIIa and FcγRIIIb and their allotypes, which confirms and expands recent studies using a limited set of glycovariants presented here ([@B24], [@B30]). Increased FcγRIII binding seems to be a general phenomenon for all IgG subclasses upon afucosylation ([@B50], [@B51]). The positive binding effects were primarily caused by the lack of fucose, which was further strengthened by additional galactose. A similar effect has been observed for neutralizing anti-HIV antibody 2G12 produced in modified plant cells which showed better FcγRIIIa binding and antibody-dependent (NK) cell-mediated viral inhibition ([@B52]).
The enhanced binding of galactosylated and afucosylated IgG was slightly weakened by addition of sialic acid, but only if a bisecting GlcNAc was present. A similar negative effect of sialylation has previously been observed for mouse FcγR by Ravetch and colleagues ([@B53]). Importantly, we showed that the enhanced FcγRIII-binding effects are directly translated into increased FcγR-mediated cellular functions. We tested this using NK cell-mediated ADCC, as NK cells are the only cell type that only express FcγRIIIa. Curiously, we observed no ADCC at all for fucosylated IgG, even at high concentrations of IgG1. Thus, ADCC activity was only observed with afucosylated IgG1. Although somewhat surprising, this phenomenon has been observed previously for anti-Rhesus-mediated ADCC ([@B54]), but also for Rituximab-mediated B cell killing ([@B27]). This suggests that the enhanced affinity afucosylation of IgG has on FcγRIIIa binding is required to cross a signaling threshold of FcγRIIIa on NK cells required for killing.
The second surprise was that no significant difference was observed between ADCC-capacity of NK cells from donors homozygous for one of the two FcγRIIIa-V/F158 allotypes, of which the V158 allele is known to have higher affinity for IgG (also confirmed here to be \~2--5×) ([@B49]). *In vitro*, this has been found result in stronger functional efficacy for the V158-variant ([@B55]--[@B57]). *In vivo*, conflicting reports have showed that individuals homozygous either the V158 or the F158 allotype show stronger cellular clearance ([@B58]--[@B61]). It should be noted that most of these studies were performed before the knowledge of FcγRIII gene being influenced by copy number variation ([@B61]). We also now know that NK cells can also express FcγRIIc or FcγRIIb in some individuals. Both these variations affect the functionality of this receptor ([@B32], [@B62], [@B63]). In this study, we eliminated both these variables by selecting donors with two copies of FcγRIIIa and without FcγRIIc-ORF, possibly explaining these discrepancies, and perhaps suggesting that the twofold to fivefold difference in affinity of IgG1 allotype is not enough to cause functional differences.
Importantly, the observed changes in FcγRIIIa-binding due to glycosylation reliably translated into functional NK cell-mediated ADCC lysis of RBC. For FcγRIIIa and FcγRIIIb it was known that absence of IgG-Fc core-fucosylation increases the affinity of interaction due to a glycan--glycan interaction between the Fc glycan and the N162-glycan uniquely found in the FcγRIII family ([@B11]). Our approach to combine this with multiple end glycan editing shows an additional layer of complexity exerted by the galactose and sialic acid. The reasons for this added effect of galactose is unknown but may very well be related to the subtle effects on quaternary structure of the Fc-domain ([@B64], [@B65]) but may also be related to differential interaction of the Fc-glycan with the N162-glycan found in FcγRIII ([@B11]).
The possible effect of the Fc-glycans on complement activity, has until now remained enigmatic. It has been proposed for a long time that agalactosylated IgG activates complement more efficiently through the lectin pathway (MBL) ([@B25]). To our knowledge these results have never been confirmed. On the contrary, we saw enhanced complement activity of all glycovariants with elevated galactose, and no evidence of MBL being activated by any of our glycoforms. These results confirm recent work also suggesting galactosylation of IgG1 to positively influence C1q binding and CDC ([@B26], [@B66]). In addition, our results clearly rule out fucosylation or bisection having an effect on complement activation, and we now show that sialylation increases the C1q-binding of galactosylated IgG. This effect of sialylation was observed on all different glycan backbones (e.g., with or without fucose, with or without bisection) which is highly suggestive that this is no artifactual finding. This is in contrast with the previously mentioned study showing that additional sialylation decreases C1q binding ([@B26]). Activation of complement is dependent on spatial arrangement of the IgG on the cell surface ([@B67]) which is likely to differ considerably between each monoclonal antibody and target, and may possibly explain the discrepancies found between our two studies. This view is supported by our observations that sialylation had limited if any effect on IgG-mediated CDC using RBC as targets, while binding to C1q of anti-TNP antibodies was enhanced by sialylated IgG on solid surfaces.
Low galactosylation level in total IgG generally correlates with disease severity of several autoimmune diseases, such as rheumatoid arthritis and multiple sclerosis ([@B13], [@B14]). While this may seem at odds with our observations at first glance, with high galactosylated IgG having elevated complement and FcγR activities, both notions are in agreement if the balance between total- and antigen-specific glycosylation is taken into account. In this way, low potential for FcγR- and C1q binding for total IgG (e.g., low galactosylation), creates a pro-inflammatory environment in which clinical manifestations can take hold as this lowers the threshold for pathogenic antibodies. Antigen-specific IgG can also potentially have different glycosylation features than total IgG as we have shown before ([@B17], [@B18], [@B21], [@B22]), and if these are more pro-inflammatory than that of total IgG, this can theoretically lead to enhanced immune activation and clinical symptoms. The knowledge obtained in the current research provides a roadmap to decipher the meaning of glycan profiles in these diseases settings.
In summary, we show here that a set of glyco-engineering techniques we recently developed ([@B31]) can be combined to quickly generate any desired IgG glycoforms to test the effect on functional capacity. Using two sets of monoclonal antibodies we generated the most extreme 20 different glycoforms possible, and examined their effect on binding to FcγR and complement, as well as their functional capacity to trigger cytotoxicity. These revealed first that the normal glycosylation changes seen in human IgG1 do not affect any other FcγR than FcγRIIIa and FcγRIIIb. Second, hypo-fucosylation and galactosylation increase binding to both human FcγRIII-variants, with a minor negative effect of sialic acid and bisecting GlcNAc. In addition, galactosylation is the primary glycan adduct that enhances C1q-binding and all downstream complement activities, including CDC. This is summarized in Figure [6](#F6){ref-type="fig"}. Collectively, this indicates that afucosylated and hyper-galactosylated IgG1 antibodies have both improved ADCC and complement-mediated activities, including complement opsonization and CDC. These properties can now be systematically implemented in new therapeutic antibodies for enhanced effector functions. Even as important, this also allows us to decipher the clinical potency of antibodies in immune responses that have tendency to have altered fucosylation and/or galactosylation ([@B17], [@B18], [@B21], [@B22]).
{#F6}
Ethics Statement {#S5}
================
Peripheral blood from anonymous, healthy volunteers was obtained with informed, written consent of all subjects, in accordance with Dutch regulations. This study was approved by the Sanquin Ethical Advisory Board in accordance with the Declaration of Helsinki.
Author Contributions {#S6}
====================
GD, TK, DW, TR, MW, and GV designed the research. GD, AB, DW, TR, MW, and GV designed the experiments. GD, LT, RP, AB, MdB, CK, SL-T, RV, and YM performed the experiments. GD, RP, AB, MdB, CK, TR, TK, MW, and GV analyzed data, GD and GV wrote the manuscript. All authors contributed to and approved the final manuscript.
Conflict of Interest Statement {#S7}
==============================
The authors declare that this study received funding from Sanquin Bloedvoorziening, a not-for-profit organization. The funder was not involved in the study design or collection, analysis, or interpretation of the data.
The authors would like to thank Ninotska Derksen, Pleuni De Heer-Ooijevaar, Prof. Dr. Rob Aalberse, and Sanne van de Bovenkamp for practical help and Prof. Dr. Ellen van der Schoot, Prof. Dr. Rob Aalberse, Sanne van de Bovenkamp, Dr. Juan J. Garcia-Vallejo, Willem Falkenburg, and Christine Bruggeman for fruitful discussions, and Prof. Dr. Ellen van der Schoot for critically reading the manuscript.
**Funding.** This study was supported by Sanquin Product and Process Development Plasma Products, 12-001, Gestur Vidarsson.
Supplementary Material {#S8}
======================
The Supplementary Material for this article can be found online at <http://journal.frontiersin.org/article/10.3389/fimmu.2017.00877/full#supplementary-material>.
######
Click here for additional data file.
[^1]: Edited by: José Mordoh, Fundación Instituto Leloir, Argentina
[^2]: Reviewed by: Raffael Nachbagauer, Icahn School of Medicine at Mount Sinai, United States; Johannes S. Gach, University of California, Irvine, United States
[^3]: Specialty section: This article was submitted to Vaccines and Molecular Therapeutics, a section of the journal Frontiers in Immunology
| {
"pile_set_name": "PubMed Central"
} |
Deception detection research has primarily focused on developing an understanding of the strategies that individuals employ when lying or telling the truth about solo crimes, and furthermore what interviewing techniques can be employed by investigators to enable individual liars to be more easily differentiated from individual truth-tellers (e.g., [@B99]; [@B17]; [@B110]). However, despite lone individuals being the focus of deception detection research, investigators are often faced with situations whereby they must interview more than one person. Group (or co-) offending (an offense committed by two or more people) is common ([@B66]) and on the rise ([@B91]). Studies suggest that co-offending accounts for between 10 and 17% of police recorded offenses with between 22 and 30% of offenders being involved in co-offending ([@B14]; [@B43]; [@B90]; [@B20]; [@B71]). However, co-offending statistics need to be interpreted with caution. This is because the data available tends to only represent those offenders who have been charged or detected, resulting in a sizeable dark figure of co-offending (e.g., a crime may be recorded as a solo event when in fact investigators are just unaware that more offenders were involved).
Co-offending is just one circumstance that would require the investigation of more than one person. Other contexts in which multiple individuals may need to be questioned are when there are multiple eyewitnesses to a specific crime or when a suspect reports an alibi witness (someone who can provide an account of the whereabouts of a suspect at a location other than the crime scene at the time the crime took place; [@B21]). Additionally, some criminal groups may consist of members who are involved only in the planning, or in the aftermath, of an offense, but not in the actual commission of the offense itself. Similar to alibi witnesses, these group members will be collaborators of the crime but not active participants. Despite having different group dynamics to multiple suspects, there is still the potential for investigators to question more than one person about the same crime/event. However, in these contexts the aim for investigators is to determine the accuracy of the information provided by the eyewitnesses, to determine whether the alibi witness is true or false, or to determine the differing roles played by the group members in the offense process.
Although investigators often have to question groups, whether it is multiple suspects, multiple eyewitnesses, or alibi witness situations, the number of deception detection studies involving groups is limited. Our literature search found only 20 studies published on the topic of verbal group deception (see **Table [1](#T1){ref-type="table"}**). All 20 studies, except one, considered multiple suspects; therefore it is groups of suspects that will be the primary focus of the current review. The aim of the current review article is to collate and discuss the research that has been conducted into interviewing groups to detect deceit and to discuss the theoretical implications of this upcoming research area. The review will consider the differences between groups and individuals, and will report the different aspects of the extant group deception studies, for example, interviewing group members separately vs. interviewing group members simultaneously. Throughout, the review gaps in the literature will be identified with suggestions for future studies. The current review will focus on the verbal indicators of deceit and therefore studies measuring physiological measures to detect deceit in groups will not be discussed (see [@B13]; [@B68], [@B67], for studies using the Concealed Information Test (CIT) to extract information from groups).
######
Overview of the group deception studies completed so far, which measure verbal cues (spanning across 11 years from 2003 to 2014).
Group deception study Theoretical principle examined Interview style Study manipulation Number of interviewees Adults or children? Published?
----------------------- -------------------------------- ----------------- --------------------------------------- ------------------------ --------------------- ------------
[@B33] Strategies Individual SUE technique 3 Adults Yes
[@B104] Strategies Individual (Un)anticipated questions 2 Adults Yes
[@B15] Consistency Individual Effect of co-planning 2 Adults Yes
[@B32] Consistency Individual SUE technique 3 Adults Yes
[@B37] Consistency Individual Repeated interviews 2 Adults Yes
[@B60] Consistency Individual Drawings 2 Adults Yes
[@B65] Consistency Individual (Un)anticipated questions: intentions 3 Adults Yes
[@B77] Consistency Individual Sketches 3 Children Yes
[@B79] Consistency Individual (Un)anticipated questions: intentions 2 and 4 Adults Yes
[@B84] Consistency Individual Response mode of observers 2 Children Yes
[@B85] Consistency Individual Repeated interviews 2 Adults Yes
[@B97] Consistency Individual (Un)anticipated questions 2 Children Yes
[@B106] Consistency Individual (Un)anticipated questions 2 Adults Yes
[@B23] Memory Collective Brief investigative interview 2 Adults Yes
[@B54] Memory Collective Timeline task 2 Adults Yes
[@B55] Memory Collective Pairs' monitoring of the interviewer 2 Adults Yes
[@B93] Memory Collective (Un)anticipated questions 2 Adults Yes
[@B105] Memory Collective (Un)anticipated questions 2 Adults Yes
[@B94] Imposing cognitive load Collective Forced turn-taking technique 2 Adults Yes
[@B69]^∗^ Verifiability approach Collective Written statements 2 Adults Yes
SUE, Strategic Use of Evidence.
∗
This study involved alibi witness scenarios as opposed to multiple suspects.
Defining A Group
================
A group can be defined as "two or more individuals who are connected to one another by social relationships" ([@B27], p. 3). Group deception studies, to date, have primarily focused on dyads, and the majority of group offenses in real-life involve dyads ([@B44]). Nevertheless, it is important to note that there is much debate within the group dynamics literature with regard to the definition of a 'group' and whether a dyad actually constitutes a group (often referred to as the Moreland-Williams debate). [@B62] argue that dyads are not groups because the number of individuals within the group matters. That is, larger groups involve different group processes. They state that dyads are different from larger groups because; (1) they form and dissolve more quickly; (2) they are more emotionally involved with one another; and (3) certain social phenomena, such as minority/majority relations, cannot be applied unless there is more than two people in the group. In contrast, [@B117] states that dyads should come under the definition of a group because group processes, such as social loafing and facilitation or ingroup/outgroup, occur regardless of whether there are two or several people in the group. Consequently, dyads are appropriate targets for investigation. [@B117] reports that it is the scientific methods employed which are important to ensure that reliable data is collected about how individuals behave when around others. If dyads do not constitute a group, nor are they individuals, then dyads will need to be studied as a field on their own. As a result of this controversial debate within the group dynamics literature, it is important that group size is considered when interpreting the findings and forming conclusions from the current review. Nevertheless, the fact that dyads are different from individuals regardless of whether they come under the term 'group,' means that the current review is still necessary and relevant for those involved in investigative interviewing and detecting deception. For the purpose of the current review and in line with the definition by [@B27] mentioned above, dyads will be thought of as a group.
Group Deception: How Are Groups Different From Individuals?
===========================================================
Although the knowledge-base surrounding detecting deceit in groups is limited, it is important to acknowledge that groups are different from individuals (e.g., in terms of shared responsibility and peer support; [@B111]). Consequently, there are numerous reasons why it is important to expand the investigative interviewing and deception detection literature to include groups. First, the characteristics of group offenders differ from the characteristics of solo offenders. For example, co-offenders are typically younger than solo offenders ([@B90]). Furthermore, it seems that groups are more strategic and view honesty and deception differently from individuals. Groups, for instance, lie more than individuals ([@B18]) and are more likely to use honesty strategically in order to maximize their own outcomes (e.g., financial payoffs; [@B87]). Groups also report more self-interest (both collective self-interest and individual self-interest) and fewer concerns about using deception in comparison to individuals ([@B18]). These differences in motives, strategies and behaviors highlight the need to understand group deception as well as individual deception. Additionally, there are specific crimes (e.g., organized crime, terrorism, drug trafficking, burglary, arson) and multiple investigative settings (e.g., immigration, airport security, border control, police stop and search) that are more likely to involve groups of offenders as opposed to lone offenders ([@B14]; [@B90]).
Second, the interviewing of groups brings a different dynamic to the interview process. This allows for different interview approaches to be applied and enables the identification of unique cues to deceit that can only emerge when groups are interviewed. A central issue that investigators face is whether to interview groups individually (i.e., interviewing of group members separately) or collectively (i.e., interviewing of group members simultaneously). Interviewing group members individually provides the potential to examine within-group consistency ([@B85]; [@B32]), which cannot be measured when interviewing a lone individual about a solo crime. Additionally, collective interviewing allows for communication and interaction cues to be examined ([@B23]; [@B105]; [@B55]; [@B94]), and these cues cannot emerge when interviewing a lone individual about a solo crime or when individually interviewing group members about a joint crime. Currently, police typically conduct individual interviews during their investigations regardless of the number of suspects. Nonetheless, there are some existing procedures where collective interviewing is employed. For example, in the UK, immigration officers occasionally use collective interviewing when attempting to uncover sham marriages ([@B50]), and police detectives often interview people in groups when making house-to-house enquiries ([@B19]). In Canada, customs officers carry out collective interviews at airports because members of the same group are deemed to have a 'similar agenda' (i.e., it is assumed that group members traveling together are traveling to the same destination for the same purpose). Thus, if only one person in the group is examined, this could result in a wasted effort or missed opportunity. That is, the interviewing of group members individually will be less time-efficient and may result in specific individuals who need to be interviewed being disregarded (personal communication with a Canadian ex-immigration officer, 12th November 2013). By extending the research agenda to include collective interviewing, research can inform on the best practices for such situations and has the potential to uncover new applied contexts where collective interviewing may be appropriate.
Third, for groups you can determine deception at a social level as well as at an individual level. In other words, interviewing groups of suspects not only allows for the additional consistency and social cues (as mentioned above) to be measured, but also allows for the measurement of established deception cues within each individual of the group (e.g., within-statement consistency, number of details, admitting lack of memory, plausibility, spontaneous corrections; [@B35]; [@B22]; [@B98], [@B99]). Therefore, by examining individual deception and group deception more cues to deceit can emerge from interviewing group members about joint events than from interviewing lone individuals about solo events. The identification of both known and novel cues to deceit from groups is important if we are to improve the poor accuracy rates typically found within the detection deception literature ([@B12]) and obtain knowledge of what cues are best to focus on in each given investigative situation.
Fourth, when multiple people need to be questioned, each member of the group loses some control over what information is provided and how each member behaves. Therefore, not only do they have to 'impression manage' themselves (i.e., regulate and control what information they say), but they also have to 'impression manage' others (i.e., attempt to regulate, predict, and control what others say). This suggests that the strategies employed by groups will differ from the strategies employed by lone individuals because group members will not be as successful as lone individuals when improvising on the spot during the interview. This will be the case regardless of whether the group members are interviewed individually or collectively. The notion of impression management can be linked to research examining the 'Prisoner's Dilemma' ([@B76]; [@B73]), especially when group members are interviewed separately about a crime. That is, each member of the group will need to make a decision about whether they can trust the other group members to cooperate with the group and not the police. Such reasoning is difficult. Therefore the added task of impression managing others implies that group situations are naturally more cognitively demanding: not only do you have to think about and remember what you have said but you also need to consider what others have said or might be saying. Research shows that increased cognitive load results in more cues to deceit ([@B100], [@B103]). Hence, cues to deceit may not only be more numerous but also strengthened in group situations.
In order for law enforcement to be competent in dealing with all forms of crime-related incidents, they need to be trained in a variety of interview techniques that enable them to question both individuals and groups. Different situations may require different interviewing techniques. It is therefore important that investigators have a wide range of all the necessary skills to be able to use the most appropriate interview strategy for the situation at hand, and to be able to recognize the cues indicative of deceit that emerge from implementing that particular interview technique. For example, if there is only one interviewer available but multiple suspects, employing a collective interviewing approach would be most convenient and time-efficient. When this approach is implemented, the focus should be on cues associated with how the interviewees interact and communicate with one another, such as verbal transitions ([@B23]) or eye contact ([@B55]).
Theoretical Implications of Detecting Deception Within Groups: What Do the Existing Deception Studies Tell Us?
==============================================================================================================
From a theoretical perspective, research on the interviewing of groups contributes to the knowledge-base within the deception literature. It adds to the consistency research (e.g., repeat vs. reconstructive hypothesis; [@B35]), compliments memory research (e.g., collaborative learning and remembering), and can be applied to the theories of co-offending (e.g., social exchange theory; [@B112]) and other group processes (e.g., group formation and leadership). Hence, the interviewing of groups about joint events not only allows for the development of new interview techniques that increase information elicitation (crucial to detecting deceit), but also allows for additional theories and concepts to be applied to deception that cannot be applied when interviewing lone individuals about solo events. In the following sections, we will review the existing deception research on group interviewing which has contributed to the literature on; counter-interrogation strategies (e.g., [@B33]), within-group consistency (e.g., [@B85]), transactive memory (e.g., [@B113]), and collective memory (e.g., [@B7]). We will also highlight theoretical positions we feel are of utmost relevance to group deception, but have yet to be empirically examined, such as group dynamics (e.g., [@B5]).
Strategies
----------
Knowledge about the strategies that truth-tellers and liars employ is important if investigators are to have insight into what truth-tellers and liars will say, and how they will behave, during an interview. If this is known, more insight can be acquired with regard to the deception cues that may arise, and most importantly, more effective and theory-based interview protocols can be developed to help improve investigators' abilities to accurately detect deceit (see [@B34] for an overview). Despite this, most research has focused on the verbal and non-verbal behaviors of truth-tellers and liars, with little research being conducted into the actual strategies that truth-tellers and liars employ to appear credible (e.g., [@B22]; [@B99]). This is particularly the case for the strategies employed by truth-telling groups and lying groups (see [@B104]; [@B33] for the only two deception studies to examine group strategies when individual interviewing is applied).
When faced with the situation of being questioned about their activities or whereabouts at the time of a crime, truth-tellers and liars (whether an individual or part of a group) have the same goal and that is to convince the interviewer of their innocence ([@B31]). Research into the differing strategies employed by truth-telling groups and lying groups, when group members are interviewed individually, demonstrates that although there are no differences in terms of the non-verbal strategies (e.g., both truth-telling groups and lying groups plan to suppress nervous behaviors), there are key differences in terms of the verbal strategies they use (i.e., how they provide information; [@B104]). This is also the case for the non-verbal and verbal strategies used by lone individuals ([@B86]).
Truth-telling groups are less likely to have a strategy than lying groups because truth-tellers believe in a 'just world' ([@B61]) and that the truth will shine through ('illusion of transparency'; [@B29]). Consequently, they rely on memory, as opposed to preparation, to provide their answers. Truth-telling groups prefer a 'tell it all' strategy that aims to provide an honest and detailed description of what actually occurred ([@B104]; [@B33]). Even when part of a group, truth-tellers do not feel the need to prepare, believing their statements will naturally be consistent with one another. The only time truth-telling groups may prepare is to run through what happened in order to remind one another of the details of the event ([@B104]).
Conversely, lying groups, who are less likely to take their credibility for granted, are more likely to prepare for an interview. They plan what to say beforehand and prepare fabricated stories that are coherent and plausible. Lying groups prepare joint alibis, preferring a 'keep it simple' strategy in order to avoid raising suspicion and to ensure consistency within and between their statements ([@B104]; [@B33]). They will prepare answers to possible questions to 'get their stories straight,' but their answers will be restrictive and vague to reduce the chances of them contradicting one another or providing incriminating evidence ([@B104]; [@B33]). These differing strategies employed by groups of truth-tellers and groups of liars results in the statements from lying groups being less detailed than the statements from truth-telling groups. However, because lying groups have planned what to say together, their statements are often as consistent as the statements of truth-telling groups ([@B37]; [@B15]).
As a result of the strategies that lying groups employ, opportunity for planning is likely to be an important moderator of credibility. Research on lone individuals has shown that liars who prepare for an interview are more difficult to distinguish from truth-tellers, than liars who do not prepare ([@B70]; [@B12]). Preparing what to say in an interview enhances liars' deceptive performance ([@B30]) by making their statements not only more consistent than truth-tellers' ([@B37]), but also more immediate and plausible ([@B15]). Consequently, the emergence of cues to deceit is reduced ([@B22]). Nevertheless, the influence of opportunity to plan has not been empirically tested in groups of interviewees, and therefore studies examining the effects of planning on lie catchers' abilities to distinguish between groups of truth-tellers and groups of liars are required.
The Al Qaeda Handbook (a terrorist training manual found in 2000 by Manchester police in England) offers further insights into the strategies of groups of suspects. Amongst other things, this handbook underscores the importance of a group security plan that all group members are to commit to. This plan emphasizes the importance of teamwork and explains how to undertake group missions, which includes practicing answers to a list of anticipated interview questions. Additionally, it covers counter-interrogation strategies. Of specific relevance to group interviewing situations are the strategies that state; (i) always stick to the cover story even when shown evidence of involvement, and (ii) if interrogators state that other group members have revealed information, just agree, but do not state any additional information. This manual therefore gives insight into the types of strategies that groups may employ in order to; (1) co-ordinate with other group members; (2) avoid revealing incriminating information; and (3) avoid deviating from the group security plan. However, it assumes that group members will be interviewed separately. Therefore, interviewing group members collectively may be an interview tactic that groups typically do not prepare for, which may increase cues to deceit.
To summarize, the fact that research shows that groups of truth-tellers and groups of liars use the same non-verbal strategies but different verbal strategies helps to explain why verbal cues are deemed more diagnostic to deceit, than non-verbal cues (which is the case for both adults and children; [@B84]; [@B99]). Furthermore, whether a lone individual or part of a group, the strategies employed seem to be very similar. However, there are likely to be key differences with regard to the strategies employed by truth-tellers and liars depending on whether they are part of a group (as suggested by the Al Qaeda Handbook), because groups will need to decide what to do if the planned strategies cannot be employed during the interview. More studies investigating the strategies that groups of truth-tellers and groups of liars employ are required. The few deception studies that have considered strategies have involved interviewing groups consisting of two or three members separately. This means that the strategies employed by larger groups have been ignored, as have the strategies employed by groups interviewed collectively. Group size is unlikely to influence the counter-interrogation strategies employed, because the theoretical principles of the strategies remain the same. However, the outcome of implementing the strategies may differ as group size increases, because the ability to remain consistent will become more challenging. Future studies will need to examine this before any conclusions can be drawn. Collective interviewing raises further questions with regard to suspect strategies. For instance, will group strategies cover the communication and interaction cues that are available during collective interviewing? Or how do group members react and alter their strategies when informed that one group member has provided information other than that agreed upon in the pre-planned story? Future research on group strategies should address such questions.
Consistency
-----------
Verbal consistency is regarded by many as an important cue to deceit. Both laymen ([@B88]) and professionals ([@B2]; [@B83]) assume that inconsistency is indicative of deceit. However, it is often overlooked that consistency comes in many forms. Single statements can be examined for within-statement consistency (the level of consistency within one statement from an individual), while repeated statements can be examined for between-statement consistency (the level of consistency between multiple statements from one individual). When groups of individuals are interviewed seperately an additional form of consistency emerges: within-group consistency (the level of consistency between statements from group members; see [@B96], for an overview of consistency cues in deception contexts). The first studies on group deception began with this cue in mind ([@B37]; [@B85]), and, considering the importance that people place on consistency as a cue to deceit, it is perhaps not surprising that the majority of studies on group deception have continued to focus on this cue (see **Table [1](#T1){ref-type="table"}**). However, it is important to note that all the group deception studies that have considered consistency as a cue to deceit have involved interviewing the group members individually. Future studies should consider measuring consistency as a possible cue to deceit in collective interviewing situations.
If the strategies that truth-telling groups and lying groups employ are taken into consideration (i.e., truth-tellers rely on memory whereas liars plan what to say), it may be problematic to assume that truth-tellers are more consistent than liars. Accordingly, the consistency of truth-telling groups is often equal to (or even weaker) than the consistency of lying groups (e.g., [@B37]; [@B85]). To understand this further, [@B35] proposed the 'repeat vs. reconstruct' hypothesis, which emphasizes that liars will attempt to *repeat* what they have previously said and truth-tellers will try to *reconstruct* what they actually experienced. When truth-tellers are asked to repeat answers, their memory restructures the event so they gain, lose, and change information over time ([@B6]), thus reducing consistency. In contrast, liars merely repeat what they originally prepared, thus promoting consistency. Although, the 'repeat vs. reconstruct' hypothesis was originally developed to measure the consistency between two statements from the same interviewees ([@B35]), later research has found that this hypothesis can also be applied when measuring the consistency between statements of multiple interviewees (e.g., [@B37]). Research suggests that there is a variation in the types of details that groups provide, with truth-telling groups focusing more on the salient aspects of an event than lying groups. Thus, when these salient aspects are compared, consistency is significantly greater for truth-telling groups compared with lying groups ([@B77]). Future research should consider whether there are specific types of details that truth-tellers reconstruct and liars repeat (e.g., salient/central details vs. non-salient/general details).
Relevant to the 'repeat vs. reconstruct' hypothesis is the 'reminiscence effect' which suggests that repeated questioning results in the recall of previously unrecalled items (referred to as commission errors; [@B72]). This effect has been found to be stronger for truthful statements than for deceptive statements ([@B37]). Therefore, in contradiction to the stereotypical belief that consistency implies truthfulness (see 'consistency heuristic' literature; [@B36]), it is clear that lie catchers need to be cautious when interpreting consistent statements as truthful and inconsistent statements as deceitful. In fact, the diagnostic value (i.e., predictive accuracy) of using consistency to judge veracity in groups is modest for both adults ([@B85] obtained overall accuracy rates of between 52.5 and 70%) and children ([@B84] obtained an overall accuracy rate of 62.5%). These modest accuracy rates are not only because lie catchers are exercising the consistency heuristic incorrectly, but also because judging consistency is a subjective task (i.e., different observers can perceive the same set of statements differently in terms of consistency; [@B36]).
The diagnostic value of the consistency cue can vary depending on the response mode used. The response mode refers to the stage at which a lie catcher makes a veracity judgment, and this can vary when you have repeated interviewing and/or groups of interviewees. Research implies that observers are more accurate at using the consistency cue when they use a step-by-step response mode (observers make a veracity judgment after seeing each interrogation) compared to when they use an end-of-sequence response mode (observers make one veracity judgment after seeing all interrogations with all group members). This is because the step-by-step response mode facilitates more effective information processing of inconsistencies ([@B85]; [@B84]). Therefore, the diagnostic value of the consistency cue improves because there is a reduction in the degree of truth bias (the predisposition for observers to judge someone as telling the truth; [@B82]). However, the differences in accuracy rates depending on response mode vary across the few studies available and sometimes only approach significance (e.g., [@B85]); hence, more research is needed into the effects of response mode on the accuracy rates of detecting deceit in groups before any conclusions can be made.
An important development in deception research in recent years is the introduction of strategic interviewing techniques ([@B101]). These techniques involve asking interview questions that play on the differing strategies of truth-tellers and liars. The most relevant method applied to within-group consistency is simply to ask unanticipated interview questions, which negate the benefit of planning for the interview ([@B106]; [@B58]). Unanticipated interview questions are designed to disrupt liars' repeat strategy, thereby reducing within-group consistency. If framed correctly truth-tellers' consistency levels are unaffected because truth-tellers can still rely on memory to answer such questions. For example, [@B106] found that, when pairs were interviewed individually, there was less agreement between the answers from lying pairs, compared to truth-telling pairs, but only for unanticipated interview questions, such as questions concerning spatial details (accuracy rates for truth-tellers and liars ranged from 60 to 80%). When anticipated interview questions were asked, no differences were found between truth-telling pairs and lying pairs. Further studies have supported these findings with adult groups ([@B60]), child groups ([@B97]; [@B77]), and when the statements are on true and false intentions, as opposed to past events ([@B79]; but see [@B65] where truth-tellers showed higher levels of consistency for both anticipated and unanticipated questions). The accuracy rates obtained for truth-tellers and liars for the within-group studies that measure the classification of participants based on veracity are impressive, ranging from approximately 60 to 100%. These studies emphasize the need for investigators to develop interview protocols that include both expected and unexpected questions if they are to improve the diagnostic value of the consistency cue and enhance the accuracy rates of lie catchers whom are detecting deceit in groups.
When specific interview techniques are employed, such as the Strategic Use of Evidence (SUE) technique ([@B40]), a further consistency cue can be measured in addition to those already mentioned: statement-evidence consistency (the degree of consistency between the suspects statements and the evidence that the interrogator holds). [@B32] illustrated that when the SUE technique was employed during individual interviews with group members, lying groups demonstrated lower levels of statement-evidence consistency, within-statement consistency, and within-group consistency, compared with truth-telling groups. Consequently, if specific interview techniques are implemented during the questioning of groups then the issues associated with the consistency heuristic can be eliminated. Future studies should explore the application of other interview techniques to the interviewing of groups as well as examining the application of the SUE technique to collective interviewing contexts whereby group members are interviewed simultaneously.
The theoretical and empirical research on within-group consistency highlights both pitfalls and opportunities. On the one hand, lie catchers should be cautioned to not simply credit consistency and discredit inconsistency. On the other hand, the unanticipated question approach seems to improve the diagnostic value of the consistency cue. However, future research should explore other cues that can be measured when applying this approach to groups of interviewees (e.g., number of details or types of details). Alternative interview techniques that increase the differences between truth-telling groups and lying groups with regard to consistency need to be considered if the diagnostic value of the consistency cue is to improve.
Memory
------
It is widely acknowledged that memory plays an important role in deception ([@B38]; [@B81]; [@B95]; [@B107]). Research on memory is central to verbal veracity assessment tools, such as Reality Monitoring (RM; [@B53]) and Criteria-Based Content Analysis (CBCA; [@B57]), while [@B8] proposition of reconstructive memory is at the heart of the 'repeat vs. reconstruct' hypothesis discussed above ([@B35]). These theories approach memory from the perspective of the individual. However, cognitive psychologists have also considered memory as a social process. These social theories of memory -- including collective memory ([@B7]; [@B74]) and transactive memory ([@B113]; [@B48]) -- can offer unique insights into group deception. Social theories of memory can be applied to group deception in situations where multiple suspects state that they were doing something *together* at the time the crime took place.
Groups influence what individuals learn and how they remember information. Collective memory (often referred to as collaborative learning, collaborative remembering, or joint recall; [@B8]; [@B24]; [@B7]; [@B39]; [@B74]; [@B10]) examines this social nature of memory by treating past experiences and events as memories shared with others ([@B7]; [@B42]; [@B74]). Specifically, it explores how group members collectively recall information together ([@B75]), and so this concept is particularly important to bear in mind when groups of interviewees are interviewed collectively about joint events.
The research investigating collective memory suggests that group collaboration can aid memory through cross-cueing (where members of the group provide cues to one another that increase recall); error-pruning (where feedback from other members of the group create discussions that make people realize their recall errors); and re-exposure (hearing other group members recall information that they themselves had forgotten; [@B78]; [@B11]; [@B74]). When groups of truth-tellers are asked to recall a shared event together, they collectively recall the information, which results in truth-telling groups exhibiting interactions and communications that cannot be unveiled when they are interviewed individually (e.g., posing questions to one another, looking at one another, continuing on from one another, correcting one another, adding information to each other's accounts, finishing each other's sentences). Collective interviewing deception studies have shown that these interactions and communications occur more frequently for truth-telling groups than for lying groups ([@B23]; [@B105]; [@B54],[@B55]; [@B93],[@B94]). This is because the lying group members are merely recalling their planned, vague, fabricated story ([@B37]; [@B85]; [@B102]).
The theory of transactive memory can also be applied to group deception research ([@B23]; [@B54]; [@B93]). This theory is concerned with how groups (and individuals) process and structure information with regard to past events. It was originally developed to examine memory processes within intimate couples ([@B113]), but has now been applied to various different forms of group relationships (including larger networks), such as team performance and knowledge management within the work place ([@B4]; [@B59]; [@B63]). The theory proposes that people in close relationships share cognition and 'think together' by knowing each other's memory expertise and treating one another as external memory aids ([@B113]). This results in a specialized transactive memory system or 'division of labor' that is greater than the total of all the individual memories ([@B115], [@B114]).
The transactive memory system is active at all three stages of memory formation: encoding, storing, and retrieving. First, when information is encoded regarding a shared experience, responsibility for information is automatically divided and shared between all members of the group, so that each person knows what they are to remember as well as what the other group members are to remember ([@B49]). Second, when information is stored, each individual within the pair has remembering responsibilities, knowing what their role is, what they are to remember, and what information the other group members have access to [@B114]. Third, retrieval of information is social and interactive as the group members communicate with one another to retrieve as much information as possible. The communication with one another and the discussion of incoming information enhances their individual recollections. [@B48] refers to the 'transaction memory search' whereby group members who have experienced a past shared event make instinctive use of their transactive memory system to increase recall by posing questions to one another to check information or find out information, cueing one another to remind each other of further information, and handing over remembering responsibility to whoever best remembers that part of the event. These interactive and communicative behaviors between the group members help one another tap into their different memory domains and trigger further information, increasing recall.
When applied to a collective interviewing context, it has been shown that the honest groups display these fundamental interactive and communicative behaviors during joint recall significantly more than the lying groups ([@B93]). Lying groups, after all, are inventing shared events. Without the shared transactive memory system for encoding, storing and retrieving information, lying group members rely on their individual cognitive abilities to create a story that makes sense and matches with what the other group members are saying ([@B41]). This makes it difficult for lying groups to illustrate the same degree of interactive and communicative behaviors as truth-telling groups. Consequently, deceptive communication from group members interviewed collectively is characterized by the absence of social and interactive behaviors as they recall their fabricated story ([@B23]; [@B105]; [@B54],[@B55]; [@B93],[@B94]), and only provide prepared answers to expected questions ([@B37]; [@B85]; [@B102]).
To summarize, reconstructive memory can help explain the differences between truth-telling groups and lying groups regardless of whether the group members are interviewed individually or collectively, whereas collective memory and transactive memory can most appropriately be applied to the context of collective interviewing. Nevertheless, all three theories highlight the important role that memory plays in the recall of information and thus the detection of deception. That is, group members recalling an actual experienced joint event will do so in a different manner to group members who are attempting to recall a fabricated joint event, and the more that is understood about these differences in recall, the more that can be learnt about the possible cues to deception that may arise from groups. Future studies that explore groups within the area of investigative interviewing and deception detection should consider memory and the effects of joint recall on cues to deceit.
Although group collaboration can aid memory, it can also hinder memory. This is because other people can act as a source of misinformation whereby people conform to what other group members are saying regardless of what they themselves actually remember ([@B64]). Additionally, memory contamination can occur whereby one group member causes other group members to remember information incorrectly ([@B28]). Whilst the memory literature suggests that collaborative groups (group members recalling information together) recall significantly more information than individual group members (each group member recalling information alone), some studies show that nominal groups (pooled individuals whereby the group members recall information individually, but details are summed so that any duplicate details are removed) recall significantly more information than collaborative groups (often referred to as *collaborative inhibition*; [@B9]; [@B116]). Consequently, this collaborative inhibition needs to be considered when interviewing groups collectively to detect deceit.
Collaborative inhibition implies that interviewing group members separately is better than interviewing group members collectively (in terms of the amount of information obtained). However, individual interviewing of groups requires more resources and time, and is not suitable for all situations where groups need to be interviewed, for example, when there is only one interviewer available but multiple suspects (e.g., during police 'stop and search' or at road border control where cars containing multiple people need to be questioned). Additionally, separating the group to be interviewed removes the ability to measure communicative and interactive cues as indicators of deceit. Future research needs to consider what technique -- interviewing group members individually or interviewing group members collectively -- leads to the most accurate recall of information and also elicits useful, and identifiable, cues to deceit. It is likely that a combination of both individual interviewing, and collective interviewing, of group members will be required if all circumstances in which groups need to be questioned are considered. If interviewing groups individually and interviewing groups collectively are implemented into practice, future studies need to determine which interviewing technique should be implemented in which contexts, and if both techniques are needed for a particular situation, then the sequence in which they should be conducted needs to be established (i.e., interview the group collectively then individually or vice versa?).
Group Dynamics
--------------
Before mentioning the key concepts behind group dynamics and how these may affect deception detection, we feel it important to briefly mention theories of co-offending. At least four co-offending theories have been proposed; (1) *group influence* (social learning and group pressure lead to co-offending; [@B3]); (2) *social selection* (offenders select each other because they share similar characteristics and interests; [@B25]); (3) *instrumental* (co-offending is easier, more profitable, and less risky than solo offending; [@B108]); (4) *social exchange* (co-offending is a social exchange whereby offenders receive material reward, e.g., payments, and immaterial reward, e.g., social acceptance, that cannot be obtained via solo offending; [@B112]).
Whilst the social exchange theory explains more of the characteristics associated with co-offending than the other three theories, none of the co-offending theories fully explain *all* the characteristics necessary to understand this type of offending. However, taken together, the theories explain: (1) why offenders choose to co-offend; (2) how co-offending takes place; (3) why there is variation between offense types in terms of the proportion of co-offending to solo offending; and (4) the instability of offending groups. An understanding of co-offending and the ways in which offenders select one another and form groups could help investigators develop more appropriate techniques for dealing with groups of offenders, particularly when it comes to establishing the best ways of interviewing these groups and determining whether they are guilty of a crime. According to social exchange theory, who is involved in a particular crime depends on which group members are available and willing ([@B112]). This means that on some occasions not all group members are involved in the offense; therefore it is important to learn how different group members behave during the investigation process depending on whether they were actually involved in committing the offense (i.e., they know the specific details), or whether they are just aware of the offense (i.e., they only know who was involved, but not any details). Group members with differing knowledge about the crime will have an impact upon the amount of information that is revealed during an interview. Consequently, it is important to develop tactics that can be used to establish who knows what within a group. However, to date, no empirical studies have considered the level of knowledge distribution throughout a group and how this can affect cues to deceit.
When investigating a crime, co-offending adds a whole new aspect to detecting deceit that is not present with solo offending, *group dynamics*. Groups emerge when multiple people work together. Each of the group members bond (labeled *group formation*; [@B5]), and as the co-offending theories illustrate, group formation is important to those offenders who are working alongside others. Groups form a structure with each member having a different role and status within the group; thus, if more can be known about how best to interview group members depending on their role or status within the group then more interview protocols can be developed to aid the detection of group deception. Unlike consistency and memory processes, group dynamics have not yet been examined in the deception detection literature, so, at present, we can only speculate about how group processes may influence deception and subsequently cues to deceit. We suggest three potential areas of group dynamics that may be relevant to deception contexts: group cohesiveness ([@B26]); roles or status levels of group members ([@B47]; [@B16]); and cultural influences ([@B45], [@B46]; [@B51]).
Group cohesiveness refers to the properties of a group that effectively bind the group members together to give the group a sense of solidarity ([@B26]). There is suggestive evidence that groups of liars may view threats to group cohesiveness as threats to the group's credibility. As such, groups of liars may be more concerned with maintaining an air of group cohesion compared to groups of truth-tellers. For instance, liars place weight on maintaining within-group consistency ([@B33]), in contrast truth-tellers in collective interviewing situations may be more likely to disrupt or disagree with group members ([@B105]). More nuanced measures of group cohesiveness could provide better cues to deceit.
The roles or levels of status that develop within a group may also be of interest. Roles facilitate group functioning, influencing how group members behave and communicate with one another. For example, those of a higher status (i.e., leaders who are deemed to be more knowledgeable and able to initiate the ideas and activities adopted by the group) will be respected more than those who are of a lower status ([@B47]; [@B16]). Consequently, group members who are of a lower status will be more reluctant to express disagreement with those of a higher status, but more willing to express disagreement with those who are of an equal or lower status to themselves. Communication and interaction cues are important when groups are interviewed collectively ([@B23]; [@B105]; [@B93]). By attending to group roles and the status of group members it may be possible to develop more accurate or even novel communication and interaction cues.
A final concept to consider is how cultural influences can impact group dynamics. Individualistic cultures (predominantly Western societies) highlight the importance of self-reliance, emphasizing individual needs before those of the group. Conversely, collectivistic cultures (predominantly Eastern societies) highlight the importance of interdependence, where the well-being of each individual is related to the success of the group. Emphasis is put on group loyalty and conformity, with the self-identity of each individual developing from the relationships and interconnectedness between all group members ([@B45], [@B46]; [@B51]). The studies that have been conducted with regard to interviewing groups to detect deceit have involved participants from individualistic cultures (e.g., UK, USA, and Sweden). It is expected that findings from group deception studies using participants who support collectivism will be different from those participants who support individualism in terms of the behaviors that they show for protecting the group. It is believed that any findings obtained will be stronger from participants in collectivistic cultures because supporting the group will be more important to them, than those in individualistic cultures. Future research should address the influence of culture on group deception.
To summarize, although co-offending theories and psychological theories associated with group dynamics have not yet been applied to deception detection, they are relevant to how group members behave when being questioned. A better understanding of group dynamics will assist investigators with the best ways of interviewing group members and as a result aid with the detection of deception amongst groups.
Future Research Ideas
=====================
The importance of studying groups (as opposed to lone individuals) is increasingly being recognized within the investigative interviewing and deception detection literature. However, until very recently, the focus was purely on interviewing group members separately (*individual interviewing*). Nowadays, studies are also being conducted into interviewing group members simultaneously (*collective interviewing*; see [@B92] for an overview of this research). Nevertheless, there is still a long way to go before a more complete understanding of the deception occurring within groups and how to detect it is established. Future research ideas have been mentioned throughout the current review; however, there are some additional issues that should be considered if the true effectiveness of detecting deception within groups is to be recognized.
First, deception studies need to consider larger groups. At present, deception studies predominantly involve dyads, with only a few studies considering larger groups (e.g., [@B32] used triads and [@B79] used quartets). It should be possible to apply the findings obtained from the already completed deception experiments using groups to future studies with more than two interviewees, because the theoretical rationale on which the already obtained findings are based (e.g., consistency heuristic, reconstructive memory, transactive memory) should remain the same regardless of group size. However, as group size increases, it is likely that more cues to deceit will be elicited, because the interview process will become more challenging (particularly for liars) as each group member will have more people to manage and correspond with.
Second, there are a number of group processes that could be explored in addition to the ones already mentioned. For example, it would be interesting to consider what would happen if individual group members do not know whether their fellow group members are also being interviewed -- how does this affect their choice of strategy and the information they disclose? The 'Prisoner's Dilemma' is similar, but the group members would instead need to make decisions about cooperating with the group when lacking the knowledge about which group members are actually being interviewed. Additionally, the order in which each group member thinks they are being interviewed could be strategically used during an interview. For example, does the amount of information produced and the elicitation of cues to deceit depend on whether the group member believes they are first, or last, to be questioned? Another idea for a further study would be to include an additional dependent variable where each group member is asked who else they think is being questioned and what information they think their fellow group members will provide.
Third, deception detection studies tend to compare truth-telling participants with lying participants who have perpetrated some kind of misdemeanor that is engineered by the researchers. In real-life, it is not necessarily this clear-cut, especially if a guilty group of suspects has completely innocent people within it. Future research should explore how mixed groups of innocent and guilty individuals behave when interviewed either individually or collectively, and what cues to deceit emerge.
Fourth, the current review has considered each of the theories associated with group deceit as if they are independent of one another. However, it would be interesting for future studies to not only consider the application of each of the individual theories to group deceit and its detection, but also to consider the application of a combination of the theories to group deceit and its detection. For example, the link between joint memory recall and group dynamics and the effect this has on the elicitation of cues to deceit.
Fifth, the vast majority of studies have focused on co-offenders when other group situations are relevant for law enforcement. To our knowledge, the study by [@B69] examining deception in alibi witness situations, is the only group deception study to consider groups other than co-offenders. For a complete understanding of group deception more research of this kind is essential. Additionally, research exploring the best way of determining the accuracy of information provided when multiple witnesses are questioned about the same crime/event would contribute to the knowledge-base on the elicitation of information from groups.
Finally, studies examining the individual interviewing of groups have focused primarily on consistency as a cue to deceit, whereas collective interviewing studies have focused primarily on social cues to deceit. Future studies should therefore measure alternative cues when group members are interviewed individually and/or collectively in order to determine what other cues to deceit can be elicited from groups to enable the correct classification of groups based on veracity.
Practical Implications
======================
Over recent years research has increased our knowledge about what is happening during suspect interviews (e.g., [@B80]; [@B109]). However, in order to inform practice, there is a need to reach beyond descriptive research that focuses on gathering facts, and set up studies that can generate more normative knowledge. By doing such research one can identify how the current approaches that are used during investigative interviewing can be improved, which can also help to identify which approaches are most effective in each given context.
Although some law enforcement personnel currently conduct collective interviews in some situations, the interview manuals and training programs typically utilized by police and other law enforcement agencies focus on the interviewing of lone individuals. Consequently, the interviewing tactics and techniques that investigators are primarily taught, such as the PEACE model ([@B1]) or Behavioral Analysis Interview (BAI; [@B52]), and the cues to deceit that they choose to measure (e.g., consistency or nervousness; [@B99]), are developed around research into lone individuals. Hence, these interview manuals and training programs have very little, if any, information about how best to interview groups and what deception cues to observe when more than one individual is being interviewed about a joint offense. As highlighted within the current review, knowledge about lone individuals cannot always translate to groups of individuals, yet co-offending occurs frequently. The current review demonstrates that by not truly understanding groups, a large number of opportunities to employ novel or existing interviewing techniques to detect deceit are being missed (e.g., the ability to apply a collective interviewing approach or how best to measure and elicit within-group consistency). Hence, by offering both theoretical and empirical insights on how to interview groups of suspects, the current review can inform policy and practice. It suggests that the framework of police interview manuals and training programs should be revised to include more specific tactics and techniques for group situations. If this occurs then investigators will have a wider range of tools and a greater understanding of what interviewing techniques and tactics to employ when they have a co-offending situation and need to detect deception occurring within groups.
Furthermore, the current review has implications that stretch beyond the traditional law enforcement context. For example, it might guide policy with respect to techniques for both collecting and assessing the reliability of human intelligence. Specifically, the recent congressional amendment---To Reaffirm the Prohibition on Torture Amendment (2015)---strictly limits the US interrogation procedures to the methods listed in the ([@B89]). The amendment also requires the regular update of the manual based on the best available scientific evidence. Currently, the Army Field Manual says little about group interviewing situations. The current review therefore acts as a first step to fill this gap and as a call for researchers to continue important work on this topic area.
Conclusion
==========
The interviewing of groups to detect deceit about joint events is different from interviewing lone individuals to detect deceit about solo events. In group situations, not only can group deception be explored but so can individual deception, resulting in the measurement of considerably more cues to deceit. This is because unique cues that cannot be explored in lone individuals can also be measured, such as within-group consistency or cues stemming from the way that group members communicate and interact with one another. Additionally, being part of a group is more cognitively demanding in itself because each group member has to think about what they say as well as what others might say. Whether group members should be interviewed individually or collectively depends on each given situation. As the current review highlights, there are benefits to both techniques and unique cues to deception that can emerge depending on the interview style employed. Overall, there is the opportunity for investigators to develop interview protocols based on group dynamics that allows for key differences between truth-telling groups and lying groups to emerge. There is currently a lack of studies exploring group deception and its detection, yet a clearer and more accurate understanding of the deception occurring within groups and the strategies groups employ would benefit criminal, security, and intelligence investigations; and thus be of value to crime prevention and policy.
Author Contributions
====================
All three authors worked together to develop the idea about the review paper and how it should be structured. ZV did the majority of the writing and put the manuscript together. ZV received frequent comments and amendments from both P-AG and EM throughout the writing process. Several meetings were held with all three authors present. The manuscript has been checked by all three authors prior to submitting.
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.
[^1]: Edited by: *Ulrich Hoffrage, University of Lausanne, Switzerland*
[^2]: Reviewed by: *Michael Timothy Bradley, University of New Brunswick, Canada; James Driskell, Florida Maxima Corporation, USA*
[^3]: This article was submitted to Cognition, a section of the journal Frontiers in Psychology
| {
"pile_set_name": "PubMed Central"
} |
Abbreviations used in this paper: bHLH, basic helix-loop-helix; dblKO, double knockout; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MRF, myogenic regulatory factor.
Introduction
============
The process of skeletal muscle differentiation is orchestrated by a family of four conserved basic helix-loop-helix (bHLH) transcription factors that are collectively known as myogenic regulatory factors (MRFs). Mice harboring single mutations of either *MyoD* or *Myf5* are viable and do not have overt muscle phenotypes, suggesting that MyoD and Myf5 have considerable overlap in their roles ([@bib6]; [@bib21]). However, either MyoD or Myf5 is required for proper myogenesis during embryogenesis because compound *MyoD*/*Myf5*-null mice lack essentially all skeletal muscle tissue at birth ([@bib22]; [@bib12]). In contrast, myogenin is important for the terminal differentiation and fusion of myoblasts into mature muscle fibers ([@bib19]; [@bib31]). MRF4 appears to have a role as a determination factor in a subset of myocytes in the early somite and as a differentiation factor in later muscle fibers ([@bib12]).
Postnatal growth and regeneration of skeletal muscle is mediated primarily by a pool of myogenic stem cells known as satellite cells, which reside adjacent to the fibers. In response to damage through injury or exercise, these satellite cells activate expression of MyoD and Myf5 and undergo numerous rounds of proliferation as myoblasts. A small number of myoblasts return to a quiescent state, thus replenishing the pool of satellite cells; the remainder continue their differentiation, fusing into existing or new myofibers and expressing myogenin and MRF4 while down-regulating Myf5. In contrast to wild-type myoblasts, *MyoD*-null myoblasts grow more quickly, show aberrant expression of muscle markers, and differentiate inefficiently ([@bib23]), a phenotype that is the cause of the regeneration deficit exhibited by compound dystrophic (*mdx*) *MyoD*-null animals ([@bib16]). Conversely, *Myf5*-null myoblasts proliferate poorly and differentiate precociously ([@bib17]). The sequence of MRF expression in activated satellite cells, in conjunction with the phenotypes of single-null myoblasts and animals, argue that MyoD and Myf5 do not have identical roles in myoblast proliferation and induction of differentiation.
MyoD and Myf5 target genes have largely been examined after the onset of differentiation and, hence, are involved in producing the enormous phenotypic shift from a proliferating myoblast to a contractile, multinucleated muscle fiber. Although it has been suggested that MyoD and Myf5 transactivation is checked in growing myoblasts (whether by degradation \[[@bib29]\], modification \[[@bib15]\], signaling \[[@bib30]; [@bib14]\], or interfering heterodimerization \[[@bib2]\]), the distinct phenotypic differences that are exhibited by growth phase *MyoD*- and *Myf5*-null myoblasts suggest that MyoD and Myf5 do have active roles in myoblasts. Indeed, [@bib35] identified *Id3* and *NP1* as target genes of MyoD under growth conditions.
Distinguishing the distinct functions of MyoD and Myf5 is complicated by their abilities to auto- and cross-activate expression from the endogenous loci ([@bib28]; [@bib5]; [@bib29]; [@bib35]). Such a circular network could account for the stabilization and irreversibility of the commitment of a cell to a myogenic fate ([@bib29]; [@bib33]). However, gene expression changes resulting from the introduction of exogenous MyoD or Myf5 could be an indirect effect that is mediated through the other MRF. Expression of MyoD in the absence of Myf5 (and vice versa) permits the comparative evaluation of each factor\'s functions in myogenic commitment.
To this end, we reintroduced MyoD or Myf5 into a *MyoD*−/*Myf5* double-null (double knockout \[dblKO\]) fibroblast cell line that was maintained in high serum growth conditions. These cell lines are normally nonmyogenic but can be converted to skeletal muscle upon the exogenous expression of MyoD or Myf5. Microarray analysis identified numerous differentially regulated genes, which were further validated by examining specific candidates using real-time PCR. A number of growth phase targets were identified, demonstrating that MyoD and Myf5 are transcriptionally active in proliferating cells. Surprisingly, we did not find unique targets, and both MRFs were able to induce the expression of these genes.
MyoD, however, was strikingly more effective at activating differentiation markers than Myf5. Additional support for functional differences between MyoD and Myf5 were obtained by using chimeric MRFs that interchanged their NH~2~-terminal, bHLH, and COOH-terminal domains. The bHLH domains (DNA binding and dimerization) are highly conserved between the two genes, whereas their NH~2~- and COOH-terminal regions are more divergent. Whereas Myf5 was inefficient at inducing differentiation gene expression, the activation of a cohort of these markers by the chimeric MRFs provided strong evidence for cooperative gene activation by the NH~2~-terminal and bHLH + COOH-terminal regions of MyoD. *MyoD*-null primary myoblasts have a greatly reduced expression of the same genes relative to wild-type myoblasts. Therefore, these data support the idea that Myf5 is biased toward myoblast proliferation, whereas MyoD promotes myoblast differentiation ([@bib23]; [@bib24]).
Results
=======
MyoD and Myf5 activate skeletal muscle genes in growth conditions
-----------------------------------------------------------------
To circumvent the potential problem of auto- and cross-activation by the primary myogenic factors ([@bib5]; [@bib29]; [@bib34]; [@bib11]), we infected clonal double-null *MyoD*−*/*−;*Myf5*−*/*− (dblKO) embryonic fibroblast lines with retrovirus expressing MyoD or Myf5 or with empty control retrovirus. With this approach, we were able to assess which genes were common targets of MyoD and Myf5 versus those that could be uniquely regulated by one or the other primary MRF. These genes could be directly activated by MyoD or Myf5 or could be indirectly activated through an intermediate transcription factor; we considered both classes to be downstream targets of the primary MRFs.
Shortly after infection of the dblKO target cells with retrovirus, positive-expressing cells were purified by FACS based on GFP fluorescence expressed from the bicistronic retroviral transcript ([Fig. 1, A and B](#fig1){ref-type="fig"}). Pools of \>10^6^ cells were then harvested for total RNA after a further 24 h of culture in high serum growth conditions. Fluorescently labeled probes generated from biological--triplicate RNA samples were hybridized to MG-U74Av2 GeneChips, each containing probesets directed at ∼6,000 genes and an additional ∼6,000 ESTs. Comparison of MyoD or Myf5 arrays with control arrays produced candidate lists that were considered to contain genes potentially regulated by MyoD or Myf5 during growth phase ([Table I](#tbl1){ref-type="table"} and Table S1, available at <http://www.jcb.org/cgi/content/full/jcb.200502101/DC1>).
{#fig1}
###### Candidate MyoD/Myf5 target genes
GeneChip (U74Av2) Real-time PCR
---------------------------------------------------------------------------------- ---------------------- ------------------- --------------- ------ ------- ------- --------
**Transcription factors**
*Myogenin* [X15784](X15784) 10.1 1.2 2.8 A 27.0 8.4
*Sine oculis--related homeobox 1* (*Six1*) [X80339](X80339) 3.8 3.5 2.2 P/A 2.6 2.2
*Inhibitor of DNA-binding 2* (*Idb2*) [AF077861](AF077861) 3.6 2.4 2.7 P − −
*Runt-related transcription factor 1* (*Runx1*) [D26532](D26532) 3.4 2.9 3.3 P/A 4.3 2.7
*Hairy and enhancer of split 6* (*Hes6*) [AW048812](AW048812) 2.7 0.9 1.1 A 1.3 1.0
*Lung carcinoma myc-related oncogene 1* (*L-myc*)[4](#tfn4){ref-type="table-fn"} [X13945](X13945) 2.3 1.5 0.8 A 4.7 2.6
**Adhesion and receptors**
*Cholinergic receptor*, *nicotinic*, *β-1* [M14537](M14537) 24.6 16.0 15.8 A 5.0 6.1
*Cholinergic receptor*, *nicotinic*, *γ* [X03818](X03818) 9.3 2.0 4.8 A 18.1 16.4
*Cholinergic receptor*, *nicotinic*, *α-1* [M17640](M17640) 8.4 2.6 2.5 A 3.2 1.9
*Cadherin 15* (*m-cadherin*)[4](#tfn4){ref-type="table-fn"} [AJ245402](AJ245402) 4.2 0.5 1.9 A 290.6 85.4
*Transmembrane 4 superfamily member 6* [AF053454](AF053454) 2.6 2.3 2.5 P/A − −
*Discoidin domain receptor family*, *1* [L57509](L57509) 2.3 1.6 2.6 A − −
**Secreted factors**
*Insulin-like growth factor--binding protein 5* [L12447](L12447) 142.6 9.8 7.0 A 123.1 89.0
*Mast cell protease 8* (*Mcpt8*) [X78545](X78545) 11.7 11.1 14.4 A 6.2 15.6
*Secreted phosphoprotein 1* (*Spp1*) [X13986](X13986) 8.3 5.7 4.6 P 11.5 18.5
**ESTs**
*RIKEN cDNA 1190002N15* [AW125453](AW125453) 5.0 1.7 2.9 P/A − −
*cDNA clone* [AW120874](AW120874) 3.2 1.3 1.9 P − −
*RIKEN cDNA 2610201A13* [AA222883](AA222883) 1.4 2.1 1.2 P − −
*cDNA clone* [AA796831](AA796831) 1.1 2.1 1.7 P − −
**Others**
*H19 fetal liver mRNA* [X58196](X58196) 35.6 6.1 13.7 P/M/A − −
*C1q/tumor necrosis factor-related protein 3* [5](#tfn5){ref-type="table-fn"} [AI315647](AI315647) 5.8 1.9 3.1 P − −
*K+ channel tetramerization domain 12* [5](#tfn5){ref-type="table-fn"} [AI842065](AI842065) 4.9 6.0 3.2 P/A 1.1 1.3
*Protein kinase inhibitor*, *α* (*Pkia*) [AW125442](AW125442) 4.6 1.6 2.1 A 1.8 1.0
*UDP-glucose ceramide glucosyltransferase* [AI853172](AI853172) 4.3 3.3 2.9 P − −
*Paternally expressed 3* (*Peg3*) [AF038939](AF038939) 3.6 1.0 1.9 P − −
*Enolase 3*, *β muscle* [X61600](X61600) 3.5 1.8 2.0 P − −
*H2B and H2A histones* [X05862](X05862) 3.3 0.6 1.7 A − −
*α-methylacyl-CoA racemase* [U89906](U89906) 3.0 2.0 1.7 P/A − −
*Ankyrin repeat domain 1* (*cardiac muscle*) [AF041847](AF041847) 3.0 3.0 3.1 P − −
*SH3 domain GRB2-like B1* (*endophilin*) [AI842874](AI842874) 2.2 1.4 2.0 P − −
*WNT1-inducible signaling pathway 1* [AF100777](AF100777) 2.2 2.3 2.6 P − −
*ADP ribosylation factor-like 6 interacting protein 5* [AW049647](AW049647) 2.1 1.5 1.6 P − −
*SKI-interacting protein* (*Skiip*) [AW046671](AW046671) 1.2 4.8 3.6 P/A 1.0 0.5
*Enabled homologue* (*Drosophila*) [D10727](D10727) 1.9 2.4 2.9 P/A − −
**Differentiation markers**
*Actin*, *α*, *cardiac* [M15501](M15501) 240.7 3.1 9.6 A 42.9 10.1
*Troponin C*, *cardiac*/*slow skeletal* [M29793](M29793) 172.8 9.3 16.8 A 1.8 1.0
*Actin*, *α-1*, *skeletal muscle* [M12347](M12347) 93.3 9.2 14.7 M/A 265.6 130.9
*Troponin T3*, *skeletal*, *fast* [L48989](L48989) 49.3 2.8 10.4 A − −
*Myosin light chain*, *phosphorylatable*, *fast skeletal* [AV290649](AV290649) 43.5 3.0 4.6 A 3850 1540.0
*Myosin*, *light polypeptide 4* [M19436](M19436) 28.5 2.1 6.8 A − −
*Troponin T2*, *cardiac* [L47600](L47600) 21.4 2.6 3.4 P/A − −
*Myosin*, *heavy polypeptide 3*, *skeletal* [M74753](M74753) 17.0 0.8 2.0 A − −
*Troponin T1*, *skeletal*, *slow* [AV213431](AV213431) 16.4 0.8 0.8 A − −
*ATPase*, *Ca^2+^ transporting*, *cardiac fast 1* [X67140](X67140) 11.0 1.1 1.9 A − −
*Troponin I*, *skeletal*, *slow 1* [AJ242874](AJ242874) 10.5 0.9 2.7 A − −
*Myosin*, *light polypeptide 1* [X12973](X12973) 9.3 0.7 1.6 A − −
*Retinoblastoma 1* [M26391](M26391) 5.5 1.6 1.5 A − −
*Ryanodine receptor 1*, *skeletal muscle* [D38216](D38216) 4.7 1.2 1.2 A − −
*Myosin-binding protein H* [U68267](U68267) 3.9 0.5 1.5 A − −
*Sarcoglycan*, *β* [AB024921](AB024921) 3.5 1.0 1.8 A − −
*Myocyte enhancer factor 2A* (*MEF2A*) [AW045443](AW045443) 3.1 1.2 1.4 P/A 0.8 0.9
*Growth arrest and DNA damage-inducible 45 α* [U00937](U00937) 2.5 0.9 1.1 P/A − −
*Cyclin-dependent kinase inhibitor 1A* (*P21*) [AW048937](AW048937) 2.2 1.1 1.0 P 1.4 1.4
Mean fold change for pairwise comparisons of MyoD/Myf5/myogenin versus GFP from log(fold) change in MAS 5.0 software (Affymetrix, Inc.).
Present/marginal/absent call from MAS 5.0.
Fold change based on ΔCt between MyoD/Myf5 versus puromycin alone real-time PCR, normalized to *GAPDH* expression.
Added after manual inspection of dataset.
*C1q* and *K+ channel* identified in current annotations; previously listed as ESTs.
MyoD expression in a *MyoD*/*Myf5* double-null background produced increases in 47 genes (including *L-myc* and *cadherin-15*), whereas only 17 genes were increased by Myf5. Of these genes, 11 were common targets of both MyoD and Myf5; eight were activated to similar degrees by both (*Chrnβ1*, *Mcpt8*, *Spp1*, *Six1*, *Runx1*, *Idb2*, *Ugcg*, and *Kctd12*), whereas the others were more strongly activated by MyoD (*IGFBP5*, *H19*, and *α-actin*). In contrast, only six genes were down-regulated by either MyoD or Myf5 (none by both), and only four had fold changes of \<−2 (*Tcf20*, −4.9; *Dlk1*, −2.9; *Tgfbr3*, −2.7; and *S100a13*, −2.0). Relaxing the stringency of the selection criteria had only a moderate effect, producing 70 increases by MyoD and 32 increases by Myf5 versus 5 and 14 decreases, respectively (Tables S1--S4, available at <http://www.jcb.org/cgi/content/full/jcb.200502101/DC1>).
One goal of our study was to identify targets that are uniquely regulated by Myf5 but not by MyoD (and vice versa). However, very few genes were suggested by the microarray data to be increased by Myf5 and not by MyoD, and real-time PCR directed at several of those transcripts, in turn, did not support them as targets (*Skiip*, [Table I](#tbl1){ref-type="table"}; *Refbp1* and *Snk*, unpublished data). In contrast, 36 genes were up-regulated by MyoD but not by Myf5. The majority of these targets (e.g., *myogenin*, *myosin heavy chain*, and *troponin-T*), however, are associated with skeletal muscle differentiation.
At least six of the identified targets are transcription factors ([Table I](#tbl1){ref-type="table"}) and may themselves regulate the expression of other genes. Foremost amongst them is myogenin, which is an MRF that is activated by MyoD immediately upon the switch to differentiation conditions ([@bib11]; [@bib4]). Therefore, we examined (by GeneChip analysis) the possibility that our candidates were activated indirectly by myogenin using myogenin retrovirus--infected dblKO cells that were prepared as for MyoD and Myf5. 50% (13/26) of myogenin targets were also downstream from MyoD and Myf5. However, for most nondifferentiation class genes, our microarrays indicated that target activation was similar between MyoD, Myf5, and myogenin treatments, which contrasts with the considerable induction of myogenin by MyoD (10-fold) compared with Myf5 (1.2-fold; [Table I](#tbl1){ref-type="table"}). This suggests that these genes are common targets of MRFs rather than being strictly dependent on myogenin.
To identify potential growth phase targets of MyoD, probable differentiation markers were removed using previous work by [@bib4]. They used an inducible MyoD--ER fusion system to examine gene expression by microarray analysis during early differentiation in low serum conditions ([@bib4]). They identified nine subsets in their data by using a clustering algorithm to find coordinate patterns of temporal regulation, including those with transient increases, early or delayed increases, and decreases in expression. Of the 571 genes identified by [@bib4], we mapped 298 to probesets on the MU74Av2 GeneChip on the basis of GenBank, Unigene, or LocusLink IDs. Of these 298, 22 genes were also found to be up-regulated by MyoD in our experiment, and an additional seven were also up-regulated by Myf5. This provided added support to our data. A portion of these genes fell within Bergstrom clusters 5 and 6 and primarily represented differentiation-specific targets such as structural genes (e.g., *myosin* and *troponin*). However, others within clusters 1--4 (early induction) and 7--9 (decreased expression through differentiation) were also seen to have increased expression in our GeneChip data and were considered as possible growth targets.
All of the genes identified in [Table I](#tbl1){ref-type="table"} are robustly expressed by proliferating wild-type primary myoblasts ([Fig. 2](#fig2){ref-type="fig"}). Most (47/53 = 89%) of these genes are decreased in *MyoD*−*/*− myoblasts, whereas the remainder (6/53) exhibit a mixture of increase/no change/decrease calls and moderate fold increases. Differentiation markers are vastly decreased (up to 140-fold for *myosin heavy chain-3* or *troponin T1*), whereas most putative growth-phase genes showed only moderate changes (e.g., approximately eightfold for *m-cadherin* or *L-myc* and ∼2.7--1.8-fold for *Six1* or *Runx1*).
{ref-type="table"} show that the majority of differentiation markers are greatly reduced relative to wild-type myoblasts (e.g., *Myh3* to *myogenin*). In contrast, genes that are regulated in growth phase by MyoD and Myf5 are reduced to a lesser degree, if at all (e.g., *Mcpt8*, *Six1*, and *Runx1*). Calls are shown for wild-type (*n* = 3) or *MyoD*−/− (*n* = 3) myoblasts. P, present; M, marginal; A, absent. Change calls are shown for nine pairwise comparisons between wild-type and *MyoD*−/− myoblasts. I, increase; MI, marginal increase; NC, no change; MD, marginal decrease; D, decrease.](200502101f2){#fig2}
Thus, in high serum conditions, MyoD and Myf5 are capable of regulating the transcription of numerous genes. The growth phase regulation of a selection of those candidates was then specifically examined.
Growth phase candidate validation by SYBR Green real-time PCR
-------------------------------------------------------------
The expression of these candidate genes was re-examined using a second set of independently derived RNA samples. This second set was produced by the infection of a distinct dblKO cell line with drug-selectable retrovirus, yielding proliferating puromycin-resistant pools of cells that were expanded for 2 wk under drug selection. SYBR Green real-time PCR was used to quantitate target gene transcript levels using PCR primers that were chosen to span at least one intron wherever possible. The specificity of the PCR was verified by denaturing curve analysis and direct sequencing of the products.
The majority of the genes that were selected as possible growth phase target genes in the GeneChip data were also found by real-time PCR to be up-regulated ([Table I](#tbl1){ref-type="table"}). The estimates of the fold up-regulation of *Six1*, *Runx1*, *L-myc*, *IGFBP5*, and *Mcpt8* were similar by either technique. *Spp1* was increased more, and *Chrnβ1* was increased slightly less as estimated by real-time PCR; nonetheless, each was significantly increased by MyoD and Myf5. The increases in *m-cadherin* and *myogenin* levels by real-time PCR were much greater than by microarray, probably as a result of the lower background of the PCR assay. Again, however, the consistency of the changes in target expression that were produced by MyoD and Myf5 contrast with the variable induction of *myogenin* by MyoD and Myf5, arguing that this is not a strictly indirect effect.
Approximately one third of the candidates that were selected for real-time PCR validation did not exhibit significant changes in the stable pool samples ([Table I](#tbl1){ref-type="table"}). These genes tended to be those that had modest fold changes in the GeneChip data (e.g., *Hes6*, *Pkia*, *Skiip*, and *p21*). Probesets for differentiation markers that exhibited larger changes by microarray (such as *cardiac troponin C* and *MEF2A*) were likely seen as a result of spontaneous differentiation in the original GeneChip samples and were minimized in these proliferating cultures. These observations support the conclusion that the stable pools lacked the spontaneously differentiating cells that were observed in our original samples.
[@bib35] used the same inducible MyoD system as [@bib4] to generate inputs for a representational difference analysis protocol. They identified *Id3* and *NP1* as growth phase targets of MyoD ([@bib35]). Their cells were maintained in a high serum growth medium during MyoD induction, suggesting that *Id3* and *NP1* are not differentiation targets that are expressed simply as a consequence of serum deprivation. Rather, they are induced in the presence of cyclohexamide, indicating that they are likely to be direct targets of MyoD that do not require intervening protein synthesis for activation.
In contrast, our GeneChip experiments did not reveal an induction of *Id3* or *NP1* in dblKO cells by either MyoD or Myf5. To assess whether this was a consequence of low sensitivity to those genes, we used SYBR Green real-time PCR to specifically examine the expression of *Id3* and *NP1* in both the GeneChip RNA samples and in RNA from drug-selected pools; however, no significant changes were detected (unpublished data). It is likely that these inconsistencies are attributable to the numerous differences in our experimental systems, including the method of MyoD expression, the type of host cell used, and possibly the presence/absence of endogenous *MyoD* and *Myf5* genes.
From this combination of GeneChip and real-time PCR analyses, we defined a validated set of growth phase targets. Notably, MyoD and Myf5 were both capable of regulating each of these genes, and none were strictly associated with one MRF. These targets were then applied to examine the differences in function between corresponding regions of MyoD and Myf5.
Association of growth phase gene activation with MyoD and Myf5 domains
----------------------------------------------------------------------
Within their bHLH regions, Myf5 and MyoD exhibit 88% identity and \>95% homology at the amino acid level. In contrast, the regions that are NH~2~ and COOH terminal to this highly conserved DNA-binding region exhibit considerably more differences in sequence and function ([@bib10]). Therefore, we hypothesized that functional differences between MyoD and Myf5 would be a consequence of their divergent NH~2~- and COOH-terminal domains rather than a result of the bHLH recognition of discrete DNA sequences. To test this, chimeric MRFs were built, interchanging the MyoD and Myf5 NH~2~- and COOH-terminal regions around their bHLH domains. This yielded six chimeric MRF genes (termed d5d, d55, 55d, 5d5, 5dd, and dd5, where d denotes the portion derived from MyoD and 5 denotes the portion derived from Myf5) in addition to the full-length wild-type MyoD and Myf5 ([Fig. 3](#fig3){ref-type="fig"} A).
{#fig3}
Retrovirus expressing one of the chimeric MRFs were introduced into dblKO fibroblasts, yielding drug-selected pools of \>10^5^ clones from which RNA and protein were extracted for analysis. The growth phase targets identified previously were then assayed within these samples by SYBR Green real-time PCR.
Both *glyceraldehyde-3-phosphate dehydrogenase* (*GAPDH*) transcript (input RNA quantity; reverse transcription) and MRF protein levels (MRF expression) were used to enable equitable comparisons by providing normalization between samples. *GAPDH* transcripts were quantified by real-time PCR. Relative protein expression levels of full-length and chimeric MRFs were derived by direct digital densitometry of Western blots probed with three primary antibodies that each recognized unique but overlapping sets of four MRFs ([Fig. 3](#fig3){ref-type="fig"} B). All of the chimeras were demonstrated to be capable of regulating the genes identified in the previous analysis, producing substantial increases in expression compared with the puromycin-alone negative control pool.
After normalization, 5dd tended to be the poorest relative activator, whereas 55d was (in most cases) the best ([Fig. 3](#fig3){ref-type="fig"} C). However, in contrast to the other samples, these two chimeras were subject to significant normalization corrections ([Fig. 3](#fig3){ref-type="fig"} B), which may have tended to overstate their effects. Nonetheless, no single region of MyoD or Myf5 strictly correlated with enhanced or decreased relative activation. In certain cases (e.g., *Chrnβ1*, *m-cadherin*, and *Spp1*), there is a tantalizing suggestion that the Myf5 bHLH domain could have greater activity than that of MyoD, although this pattern is not borne out amongst the others. Chimeras dd5 and d55 have a greater effect than MyoD or d5d. This might suggest that an interaction between the flanking portions of MyoD acts to suppress transactivating activity, which is an effect that is disrupted when a portion of Myf5 is substituted. A similar effect was observed in deletion studies of MyoD ([@bib34]). Thus, for growth-phase genes, the corresponding domains of MyoD and Myf5 are otherwise interchangeable, and differences emerge only with respect to the degree of target induction.
The MyoD NH~2~- and COOH-terminal domains cooperate to induce differentiation
-----------------------------------------------------------------------------
A significant number of potential MyoD target genes that were identified by GeneChip analysis were differentiation markers. Wild-type MyoD and Myf5 and chimeras were expressed in dblKO cells using retrovirus and were maintained for 3 d in growth conditions before harvest. The proportion of infected cells was very similar between pools (see [Fig. 5](#fig5){ref-type="fig"} A). Differentiation marker expression was examined by real-time RT-PCR ([Fig. 4](#fig4){ref-type="fig"} A). This set of vectors included a COOH-terminal FLAG epitope tag that allowed for the normalization of gene expression against MRF protein levels ([Fig. 4](#fig4){ref-type="fig"} B). Two growth phase markers (*Chrnβ1* and *Runx1*, identified in the aforementioned GeneChip experiment) showed little relative difference between MyoD, Myf5, and the chimeras. In concordance with the GeneChip results, however, the expression of MyoD produced a considerable activation of genes such as the *cholinergic receptor* α and γ subunits, *myogenin*, *α-actin*, *myosin*, and *troponin* ([Fig. 4](#fig4){ref-type="fig"}; [Table II](#tbl2){ref-type="table"} shows unnormalized changes vs. empty vector controls). In comparison, the level of Myf5 induction of these genes was moderate relative to the empty vector control.
{#fig4}
######
**Fold activation of myogenic genes by real-time PCR** [1](#tfn6){ref-type="table-fn"}
Gene MyoD d5d 5dd 55d dd5 d55 5d5 Myf5 GFP
------------ -------- -------- ------- ------- ------- ------- ------ ------ -----
Actc1 89.2 68.5 6.5 7.8 8.9 7.6 2.8 2.5 1.0
*Myogenin* 145.0 114.8 17.4 14.5 26.2 18.4 3.7 7.4 1.0
*Mylpf* 2645.3 2727.9 249.2 356.0 211.1 234.3 62.4 81.8 1.0
*Chrng* 48.5 45.3 16.9 24.4 8.0 8.0 2.3 5.2 1.0
*Acta1* 35.4 21.8 7.7 16.1 5.7 10.8 6.9 5.6 1.0
*ChrnA1* 8.7 10.0 3.1 3.6 1.8 2.9 1.7 1.9 1.0
I*GFBP5* 127.4 104.0 104.0 91.4 74.2 35.9 22.8 23.0 1.0
*Tncc* 5.1 5.1 1.6 1.8 1.4 1.3 1.6 1.3 1.0
*Chrnb1* 11.7 14.9 10.2 11.2 6.7 8.8 7.1 7.7 1.0
*Runx1* 2.4 3.4 2.6 2.3 2.8 2.0 1.9 1.8 1.0
Relative to GFP and normalized to *GAPDH* but not to FLAG protein.
Substitution of the MyoD COOH terminus into Myf5 (55d) was not sufficient in growth phase to cause differentiation and could produce only modest increases in the expression of numerous differentiation markers. Replacement with the MyoD NH~2~ terminus alone (d55) also resulted in just moderate gene induction relative to full-length Myf5. Importantly, the concurrent presence of MyoD NH~2~- and COOH-terminal regions (d5d) resulted in activity approaching that of full-length MyoD.
The same pools of dblKO cells expressing one of the wild-type or chimeric MRFs were challenged to differentiate under reduced serum conditions. When the MyoD COOH terminus was present (MyoD; d5d, 5dd, and 55d), there was enhanced differentiation compared with the corresponding Myf5 region (Myf5; dd5, d55, and 5d5; [Fig. 5](#fig5){ref-type="fig"} B). Deletion of the putative cdk4-interacting domain ([@bib36]) from the COOH-terminal region of full-length MyoD or insertion of this region into the corresponding location of Myf5 does not significantly affect these results based on protein (myosin heavy chain immunostaining) or RNA (real-time PCR) markers (Punch, V., personal communication). The MyoD NH~2~ terminus also had a noticeable effect on differentiation ([Fig. 5](#fig5){ref-type="fig"} B, compare MyoD with 5dd, d5d with 55d, dd5 with 5d5, and d55 with Myf5). The MyoD bHLH domain, which has previously been connected to cell cycle arrest ([@bib7]; [@bib27]), enhanced differentiation when combined with the MyoD COOH-terminal region ([Fig. 5](#fig5){ref-type="fig"} B, compare MyoD with d5d and 5dd with 55d) but had little effect otherwise. Overall, the combined NH~2~- and COOH-terminal portions of MyoD (MyoD and d5d) were most effective in producing robust differentiation, whereas the two MRFs lacking both (Myf5 and 5d5) were the poorest. Thus, in both growth and differentiation conditions, the MyoD NH~2~- and COOH-terminal regions cooperate to strongly activate the myogenic differentiation program.
{#fig5}
Discussion
==========
Previous work has suggested that there are unique roles for MyoD and Myf5 in adult myogenesis ([@bib16]; [@bib23]; [@bib17]). To investigate the possibility that these phenotypes are a result of the activation of unique target genes by each of these transcription factors, we conducted genome-wide surveys of gene expression changes in response to MRF expression. The embryonic fibroblast cells that were used are normally nonmyogenic but can be converted to myoblasts by ectopic MRF expression. Importantly, they were derived in a *MyoD*−*/*−;*Myf5*−*/*− background, precluding cross-activation between MyoD and Myf5.
Our microarray data indicates that MyoD expression induces more myogenic genes than Myf5, often to a greater degree. Comparison of our data with a previous study ([@bib4]), however, demonstrated that the majority of these MyoD-regulated targets are markers of differentiated skeletal muscle. Thus, a unique function of MyoD (vs. Myf5) is a strong ability to induce differentiation. *MyoD*-null primary myoblasts have vastly reduced levels of differentiation marker gene expression ([Fig. 2](#fig2){ref-type="fig"}), which is consistent with the phenotypic data on single-null primary myoblast cultures; in the absence of MyoD, the myoblasts proliferate well and differentiate poorly ([@bib23]), whereas in the absence of Myf5, the opposite is true ([@bib17]). The moderate reduction of levels of growth-phase genes in *MyoD*-null myoblasts likely indicates a preference for activation by MyoD that Myf5 compensates for only partially.
A previous study by [@bib25] used representation difference analysis to identify potential satellite cell markers and MyoD target genes using primary myoblasts from *MyoD*−*/*− mice as a model for early satellite cell activation. The decreased expression of many genes (including *Chrnγ*, *Chrnα1*, *myogenin*, *troponin T1*, *H19*, and *Peg3*) was seen ([@bib25]); these genes were up-regulated by MyoD expression in our experiments, supporting the contention that MyoD regulates these genes and promotes differentiation. However, [@bib25] also showed that levels of numerous genes (e.g., *k-cadherin*, *integrin-α7*, *Plgf*, *VCAM1*, *Igsf4a*, *Tcrαv13*, *laminin α5*, *neuritin-1*, and *Klra18*) are increased in *MyoD*−*/*− myoblasts, which express considerable Myf5. None of these genes was identified in the current study as a Myf5 target. Thus, although MyoD and Myf5 have critical roles in the myogenic program, other transcription factors (e.g., Pax3 or Pax7) are likely needed to produce the full spectrum of normal myoblast gene expression.
We initially focused on the genes that were activated in the proliferating population by pruning our microarray results of differentiation genes. To assist in this, we leveraged the work of [@bib4], who used an inducible MyoD in differentiating dblKO cells. Their identified target genes were those induced by serum deprivation, growth arrest, and MyoD activity in the context of those conditions. However, their data also included genes that are induced immediately before growth arrest but remain up-regulated either as a result of a failure of their RNA levels to decay to baseline during the early times of the experiment or because expression continues in differentiation. Thus, we identified a set of genes that were growth phase targets of MyoD and Myf5. Furthermore, real-time PCR validation confirmed that they were not unique targets but, rather, that they all could be induced by either of these MRFs.
These similarities in function are perhaps not surprising. Both *MyoD*- and *Myf5*-null mice possess grossly normal skeletal muscle, indicating that there is redundancy and compensation for the loss of either factor during development. Indeed, only with the concurrent loss of MRF4 activity does the defect in skeletal myogenesis become fully penetrant ([@bib12]). Therefore, rather than MyoD and Myf5 each having a unique set of target genes that they are responsible for activating, it is reasonable that these two factors are capable of regulating the same downstream targets. Differences in MyoD and Myf5 function in proliferating cells may instead vary in the strength of their effect on similar sets of genes. Whether MyoD and Myf5 function similarly during differentiation is a distinct question.
Previous work by [@bib35] suggested that *Id3* and *NP1* were both activated by MyoD in myoblast growth phase. Although our data does not show this, the differences between our methods and theirs are considerable and are more than adequate to explain the discrepancy. The cell types in which the experiments were conducted are a primary example: [@bib35] used 10t1/2 fibroblasts (genetically wild type), whereas our fibroblasts were derived from *MyoD*/*Myf5* compound-mutant animals (it is also likely that our cells have disrupted MRF4 function; [@bib12]). It is quite conceivable that the presence of other myogenic factors could be capable of modulating MyoD activity.
An important conclusion to be drawn from these data is that both MyoD and Myf5 are active transcription factors in proliferating myoblasts (previously shown for MyoD by [@bib35]). Rather than passively awaiting a differentiation signal, these two factors induce the expression of myoblast growth-phase genes. MyoD and Myf5, therefore, act not only to commit the cell to a myoblast identity but also to prepare the myoblast for skeletal muscle differentiation.
The expression of MyoD or Myf5 led to the down-regulation of only a very small number of genes. This could indicate that they act primarily as transcriptional activators in high serum conditions. However, our experimental system involved converting a fibroblast cell type to the myogenic lineage and, thus, was incapable of detecting MRF-mediated suppression of any genes that were not initially expressed in the control cells.
Myogenin is not usually found in proliferating myoblasts and is induced at the onset of differentiation ([@bib11]; [@bib32]; [@bib4]). The nondifferentiation targets that were assayed (e.g., *Six1*, *Runx1*, *Mcpt8*, and *Spp1*) were induced to similar levels by either MyoD or Myf5, whereas *myogenin* was induced strongly by MyoD (10-fold) but only weakly by Myf5 (1.2-fold). Thus, although myogenin can activate similar targets to MyoD ([@bib35]) and Myf5 ([Table I](#tbl1){ref-type="table"}), indirect regulation through myogenin is not dominant. The overexpression of myogenin alone was much less effective than MyoD at inducing differentiation marker transcripts, also demonstrating that MyoD has distinct functions that are not simply consequences of myogenin induction.
The high level of structural conservation between MyoD and Myf5, particularly in the bHLH domain, provides a strong rationale for our observation that there are few, if any, unique growth phase target genes of MyoD or Myf5 (this might be extended to myogenin---the conserved bHLH domain may allow the induction of growth targets, although myogenin would not usually be found in growing cells). It has been shown previously that just three differing amino acids between the MyoD and E12 bHLH regions encode specificity for the myogenic program ([@bib8]). However, our results suggest that the three nonconservative and five conservative amino acid substitutions between MyoD and Myf5 bHLH domains are not likely to be involved in overt target gene selection, although they could potentially provide sufficient variation for more subtle types of regulation.
Chimeric MRFs, in which the NH~2~-, bHLH-, and COOH-terminal domains of MyoD and Myf5 were interchanged, were used to explore which portions of MyoD might provide for enhanced myogenic differentiation. Strikingly, only the d5d chimera approached MyoD levels of differentiation gene expression, whereas the other chimeras, Myf5, and myogenin were many times less effective. The MyoD NH~2~- and COOH-terminal regions had much less activity when separated, thus indicating a functional interaction between them.
In contrast to growth conditions ([Fig. 4](#fig4){ref-type="fig"}), differentiation conditions allowed the MyoD NH~2~- and COOH-terminal regions to function independently ([Fig. 5, B and C](#fig5){ref-type="fig"}), with the concurrent presence of both producing the greatest differentiation. Therefore, each region is effective when growth signals are naturally reduced under low serum conditions. However, cooperation between these two MyoD regions was required to overcome the strong growth signals that were provided by high serum medium. Full-length MyoD, unlike Myf5, could, therefore, bias a cell toward differentiation by sensitizing it to a moderate reduction in growth signals.
Interestingly, the MyoD bHLH domain enhanced differentiation at the molecular and phenotypic levels ([Figs. 4](#fig4){ref-type="fig"} and [5](#fig5){ref-type="fig"}), but only when combined with the MyoD COOH terminus. Despite its exceptional similarity with the Myf5 bHLH, the MyoD bHLH domain functionally differs in its interaction with the MyoD COOH terminus. However, this MyoD bHLH effect modulates gene expression levels rather than target gene selection ([Fig. 4](#fig4){ref-type="fig"}).
The NH~2~-terminal activity is likely to be the MyoD transcriptional activation domain, which was previously mapped between amino acids 3--56 ([@bib34]). It is interesting to speculate that the COOH-terminal function is provided by the chromatin remodeling amphipathic α-helix (helix III) domain that was previously described by [@bib3]. Whereas MyoD can induce muscle marker expression at repressed loci, their studies demonstrated that a similar motif in myogenin was ineffective at this task. This α-helix motif is conserved in Myf5 ([@bib10]), but Myf5 is similar to myogenin in being less effective than MyoD at initiating muscle gene expression. Thus, MyoD might be more efficient than Myf5 for inducing differentiation because of a greater ability to remodel chromatin at lineage-restricted loci. A similar domain appears in MRF4 ([@bib20]; [@bib3]) and could mediate the residual skeletal myogenesis that is found in the absence of MyoD and Myf5 ([@bib12]). Therefore, MyoD may use its COOH-terminal chromatin-remodeling domain to provide access to silenced muscle genes for the NH~2~-terminal activation domain. In support of this, the substitution of either portion alone into Myf5 produced only moderate increases in gene expression, whereas together they strongly activated a variety of differentiation markers ([Fig. 4](#fig4){ref-type="fig"}).
If MyoD is inclined to promote differentiation, whereas Myf5 activates growth (but not differentiation) targets, then it is consistent with their cell cycle regulation in growing myoblasts. MyoD levels peak at the differentiation checkpoint in G~1~ of the cell cycle, whereas Myf5 levels are high in S/G~2~ and G~0~ in association with proliferation and a failure to differentiate ([@bib13]). This complementary pattern of expression allows for the maintenance of expression of growth-phase genes and myogenic identity while ensuring that high levels of MyoD occur only at a cell cycle point that is appropriate for differentiation.
That MyoD and Myf5 activate the same downstream target genes, but to differing degrees, does much to explain the phenotypes observed in single knockout mice and cells. Our data reinforce the partitioning of myogenic factors into primary and secondary MRFs and, furthermore, add support to the concept of a role for Myf5 in myoblast proliferation versus MyoD instigating myogenic differentiation. MyoD\'s ability to activate differentiation marker expression despite the presence of high levels of serum is a distinguishing biochemical function that is not found in Myf5 and involves the cooperation of separate domains of MyoD. These regions may allow MyoD to interact with coactivators for which Myf5 has much less affinity. In the future, studies to further understand the distinct roles played by MyoD and Myf5 in growing and differentiating myoblasts, as well as the structural constraints and intermolecular interactions upon which those roles are built, will be essential to our understanding of the mechanism by which damaged skeletal muscle is efficiently regenerated.
Materials and methods
=====================
Cell culture
------------
*MyoD*-null/*Myf5*-null dblKO mouse embryo fibroblasts were isolated, and clonal lines were selected and expanded (designated 2C5/7 and 4C5/2). Fibroblasts were cultured in subconfluent conditions in growth medium of DME supplemented with 10% FCS and 1% penicillin/streptomycin. Proliferating primary myoblasts from wild-type and *MyoD*−*/*− mice were isolated and cultured as described previously ([@bib16]; [@bib23]).
*Myf5*, *MyoD*, chimeric MRFs, or myogenin were introduced into dblKO cells using retrovirus based on the three-plasmid HIT system (provided by V. Sartorelli, National Institutes of Health, Bethesda, MD; [@bib26]), including expression plasmids based on the pHAN backbone (with puromycin resistance driven from a distinct SV40 promoter) or a modified pHAN backbone in which the puromycin cassette was replaced by an internal ribosomal entry site and humanized Renilla GFP (IRES-hrGFP) sequence (Stratagene). MRFs that were expressed using the latter retroviral plasmid were COOH-terminally tagged with a 3× FLAG epitope. Empty control virus expressed only puromycin resistance or hrGRP. Retrovirus was produced by calcium phosphate transient cotransfection of *gag-pol* (pHIT60), *env* (pHIT456), and expression (pHAN) plasmids into 293T cells; virus-containing medium was harvested 48 h from the beginning of the transfection and was filtered through a 0.45-μm syringe filter (Millipore). Cells were infected overnight with diluted, filtered viral supernatant plus 8 μg/ml polybrene (hexadimethrine bromide; Sigma-Aldrich). Drug selection, where appropriate, was conducted with 1 μg/ml puromycin (Sigma-Aldrich) in growth medium for at least 1 wk; uninfected controls were obliterated after 4--5 d of selection. Differentiation was induced by replacing growth medium with DME + 2% horse serum (GIBCO BRL) and 1% penicillin/streptomycin.
FACS
----
3 d after infection with hrGFP-expressing retrovirus, pools of cells were trypsinized, centrifuged, resuspended in PBS + 5% FCS, filtered through a MACS filter (Miltenyi Biotec) to remove aggregates, and placed on ice. Cell sorting was performed using a MoFlo sorter (DakoCytomation), with gating on hrGFP+ populations set by a comparison with an uninfected negative control pool. A small portion of the sorted cells was reanalyzed to assess purity, and the remainder was replated in growth medium for an additional 24 h before harvesting for total RNA.
RNA and protein isolation
-------------------------
Total RNA was purified using the RNeasy Mini Kit (QIAGEN) according to the manufacturer\'s instructions and quantitated by OD~260~ or by RiboGreen (Invitrogen). RNA samples used for microarray analysis were ethanol precipitated in order to reach a minimum concentration of 2 μg/ml. Protein extracts were made by lysis of pelleted cells in radioimmunoprecipitation assay (RIPA) lysis buffer supplemented with MiniComplete protease inhibitors (Roche).
Microarray analysis
-------------------
Triplicate RNA samples for microarray analysis were submitted to the Ottawa Genome Centre for hybridization to MG-U74Av2 GeneChips (Affymetrix, Inc.). Manufacturer\'s quality controls were verified by the Centre. Raw data files were processed with MicroArray Suite (MAS 5.0; Affymetrix, Inc.) using the statistical algorithm (Affymetrix, Inc.) to derive signals and present/marginal/absent calls for each sample; all possible pairwise comparisons were also performed between experimental and control replicates, producing log(fold) ratio estimates of change and increase/no change/decrease calls. Processed results were then exported from MAS 5.0 and imported into Excel and Access (Microsoft) for further manipulation. Probesets showing consistent statistically significant changes between MRF and control samples were screened for detectable signal values (present/absent calls) and absolute log(fold) changes of at least one. Probeset annotations were obtained from <http://www.affymetrix.com/analysis/index.affx>. Microarray data is available from StemBase (<http://www.scgp.ca:8080/StemBase/>; Ontario Genomics Innovation Centre; [@bib18]) under experiments E223 (samples S361-4) and E59 (S78-9) and from the National Center for Biotechnology Information\'s Gene Expression Omnibus (<http://www.ncbi.nlm.nih.gov/geo/>; [@bib9]; [@bib1]) under series accession no. GSE3245 (GSM73053...64) and GSE3244 (GSM73026, -29, -32, -35, -38, and -41).
Real-time PCR
-------------
RNA samples were reverse transcribed using random hexamer primers that were included in the RNA PCR Core Kit (PerkinElmer) according to the manufacturer\'s instructions. Reverse transcription reactions were diluted (1:10 in [Fig. 3](#fig3){ref-type="fig"} C and 2:5 in [Fig. 4](#fig4){ref-type="fig"} A) with 10 mM Tris, pH 8.0, yielding master samples of reverse-transcribed products from which related PCR reactions were drawn. Real-time PCR reactions included the following: 2 μl of diluted reverse transcription product, 10 μl of 2× iQ SYBR Green SuperMix (Bio-Rad Laboratories), 30 nM ROX passive reference dye (Stratagene), and 50 nM of each forward and reverse PCR primer. Real-time data was gathered using a system (MX4000; Stratagene) over 40 cycles (30 s at 94°C, 60 s at 58°C, and 30 s at 72°C) followed by a denaturation curve from 54 to 94°C in 30-s increments of 0.5°C to ensure amplification specificity. C~t~ values were calculated with the MX4000 software by using moving window averaging and an adaptive baseline. Fold changes, other calculations, and chart plotting were performed in Excel. A PCR efficiency of 85% was assumed. Primer sequences can be found in Table S2.
Chimeric MRF construction
-------------------------
Chimeric *MyoD*/*Myf5* MRFs were created by using an overlapping PCR approach to seamlessly fuse the NH~2~- and COOH-terminal regions to the central bHLH domain ([Fig. 3](#fig3){ref-type="fig"} A). The 318 aa of MyoD (available from GenBank/EMBL/DDBJ under accession no. [NM_010866](NM_010866)) were divided into aa 1--96 (NH~2~ terminus), 97--161 (bHLH), and 162--318 (COOH terminus). The 255 aa of Myf5 (available from GenBank/EMBL/DDBJ under accession no. [NM_008656](NM_008656)) were divided into aa 1--70 (NH~2~ terminus), 71--135 (bHLH), and 136--255 (COOH terminus). Each PCR product was cloned into an appropriate expression plasmid and verified by sequencing (Applied Biosystems).
Immunocytochemistry and Western blot analysis
---------------------------------------------
Primary antibodies that were used are listed as follows: mouse anti-FLAG (M2 and M5; Sigma-Aldrich), rabbit anti-Myf5 (C-20; Santa Cruz Biotechnology, Inc.), rabbit anti-MyoD (C-20; Santa Cruz Biotechnology, Inc.), mouse anti-MyoD (5.2F; Sigma-Aldrich), mouse antimyogenin (F5D) and mouse anti-myosin heavy chain (MF20; hybridoma supernatants), mouse antidesmin (D33; DakoCytomation), and mouse anti--α-tubulin (Sigma-Aldrich).
For immunostaining, cells were fixed with 4% PFA, permeabilized with Triton X-100, and blocked with 5% normal goat serum in PBS. Primary and secondary antibodies were applied in 5% goat serum--PBS. Secondary detection used appropriate fluorescein- or rhodamine-conjugated antibodies (Chemicon). 0.25 μg/ml DAPI was included in a final wash step to highlight nuclei. Samples were mounted in fluorescence mounting medium (DakoCytomation), coverslipped, and imaged using a microscope (Axiophot 2; Carl Zeiss MicroImaging, Inc.), a 10× NA 0.30 plan-Neofluar (Ph1; ∝ /0.17) or 20× NA 0.75 plan Apochromat (∝ /0.17) objective, and a digital camera (Axiocam; Carl Zeiss MicroImaging, Inc.). Digital images were captured by using Axiovision (Carl Zeiss MicroImaging, Inc.) and were processed with Photoshop (Adobe). Enumeration was assisted by ImageJ software (<http://rsb.info.nih.gov/ij/>).
Western blots were made by electroblotting standard SDS-PAGE gels onto Immobilon-P membranes. Membranes were blocked in 5% skimmed milk in PBS; secondary detection was performed with an appropriate HRP-conjugated antibody (Bio-Rad Laboratories) visualized by ECL (GE Healthcare). For densitometry, blots were digitally imaged with a 16-bit GeneGnome chemiluminescence gel-doc system (Syngene), and bands were quantified with the bundled GeneTools software (Syngene) except parts of [Fig. 3](#fig3){ref-type="fig"} B, for which the MyoD (C-20), Myf5 (C-20), and corresponding tubulin (not depicted) blots were exposed to film, scanned, and analyzed by using ImageJ software.
Online supplemental material
----------------------------
Tables S1--S4 show the expansion of [Table I](#tbl1){ref-type="table"}, where candidate gene lists were derived by testing for consistent pairwise increase/decrease changes (at least six of nine) as well as minimum mean threshold log(fold) changes of ±1 (twofold). Increases for MyoD, increases for Myf5, decreases for MyoD, and decreases for Myf5 correspond to Tables S1, S2, S3, and S4, respectively. Table S5 shows the sequences and targets for primers that were used for SYBR Green real-time PCR. Online supplemental material is available at <http://www.jcb.org/cgi/content/full/jcb.200502101/DC1>.
We thank Vince Punch for work on MyoD(cdk4−) and Myf5(cdk4+) constructs, Dr. V. Sartorelli for providing retroviral expression plasmids, and Mark Gillespie and Mike Huh for careful reading of the manuscript.
M.A. Rudnicki holds the Canada Research Chair (CRC) in Molecular Genetics and is a Howard Hughes Medical Institute (HHMI) International Scholar. This work was supported by grants to M.A. Rudnicki from the Muscular Dystrophy Association, the National Institutes of Health, the Canadian Institutes of Health Research, the HHMI, and the CRC Program.
[^1]: Correspondence to Michael A. Rudnicki: <mrudnicki@ohri.ca>
| {
"pile_set_name": "PubMed Central"
} |
The authors wish to make the following corrections to their paper \[[@B1-ijms-19-02737]\]: We have recently been made aware of some errors and omissions in the following paragraph of our recent paper.
Section 2.2.5, lines 2--6 currently read as follows:
A structure-selective DNA binding of wild-type and G245S mutant p53 proteins on the intermolecular triplex (dT)50.(dA)50.(dT)50 has been described \[131\]. Binding of wild-type p53 on plasmid DNA containing triplex-forming sequence demonstrated that p53 prefers a superhelical form of plasmid with triplex extrusion. The use of antibodies to the p53 N- and C-terminal domains showed that the C-terminus of p53 probably plays a key role in triplex TAT binding \[131\].
To set the scientific record straight, we would like to replace this with the following:
Structure-selective DNA binding of wild-type p53 protein on the intermolecular triplex (dT)50.(dA)50.(dT)50 has been described \[131\]. Binding of wild-type p53 on plasmid DNA containing triplex-forming sequence demonstrated that p53 prefers a superhelical form of plasmid with triplex extrusion. The use of antibodies to the p53 N- and C-terminal domains showed that the C-terminus of p53 probably plays a key role in triplex TAT binding \[131\].
The original Reference 131 has been deleted, and all the reference numbers after it have been changed accordingly. References 132--155 are now numbered 131--154.
The authors would like to apologize for any inconvenience caused to the readers by these changes. The changes do not affect the scientific results. The manuscript will be updated and the original will remain online on the article webpage, with a reference to this Correction.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Human thought involves the building of mental representations by integrating external and previously stored information, and their manipulation in a cognitive space: working memory (WM; Baddeley, [@B4]). A prime characteristic of thinking is hence its abstract nature. It always requires people to construct specific representations from the perceptual features of stimuli and individuals\' knowledge and goals (Johnson-Laird, [@B39]). Moreover, higher order thinking abilities such as complex text comprehension, reasoning, and meaningful school learning are sequential. They consist of diverse component subtasks and demand that people keep their attention focused throughout the entire process. Besides the initial construction of representations, higher cognitive tasks require individuals to keep the goal of the task in mind, to shift from one sub-task to the next, to update representations by activating Long Term Memory (LTM) information, and to inhibit and discard irrelevant processes and responses. The fulfillment of these complex cognitive tasks demands that people activate all their WM resources in a controlled and supervised way (see, García Madruga et al., [@B26]).
The important role that higher order thinking abilities such as reasoning plays in knowledge acquisition and attainment at school has been addressed mainly from the perspective of the relationship between intelligence and academic performance (see, e.g., Sternberg et al., [@B71]; Deary et al., [@B20]; Vock et al., [@B79]). Classical psychometric theories have frequently defined intelligence, particularly fluid intelligence, as a capacity based on abstract reasoning (see Thurstone, [@B75]; Sternberg, [@B72]). Diverse authors have maintained that a crucial component of fluid intelligence is relational reasoning (see Cattell, [@B14]; Dumas et al., [@B21]), that is, the ability to identify and integrate the relationships between multiple mental representations. As a number of studies have shown, intelligence is probably the best single predictor of academic achievement in diverse subjects (Kuncel et al., [@B45]; Deary et al., [@B20]), particularly in mathematics-science subgroup of school subjects (Roth et al., [@B63]). The correlations found between fluid intelligence tests and academic achievement measured by means of scholastic tests are around 0.5 or higher (see, e.g., Neisser et al., [@B51]; Ones et al., [@B55]; Colom and Flores-Mendoza, [@B17]; Deary et al., [@B20]), whereas when the academic achievement measures are based on teacher grades the correlations tend to be around 0.5 or lower (see Soares et al., [@B68]). This evidence is in accordance with the results of studies that showed that increases in reasoning abilities were also accompanied by improved learning of classroom subject matter (Klauer and Phye, [@B43]). A possible explanation would be that fluid intelligence is associated with reasoning abilities (both inductive and deductive) involved in understanding and solving novel complex problems (Greiff and Neubert, [@B33]).
There is a relatively new theoretical approach on thinking and reasoning in which WM plays a crucial role: dual-process theories (see Sloman, [@B67]; Stanovich, [@B70]; Evans, [@B22]; Kahneman, [@B40]; Evans and Stanovich, [@B24]). These theories postulate the existence of two different types of thinking: System 1 (intuitive) and System 2 (deliberative). System 1 is fast, unconscious, associative and not dependent on WM. System 1 allows individuals to quickly access intuitive responses that can be valid but also a source of pervasive mistakes. On the other hand, System 2 is slow, conscious, controlled and strongly linked to reasoners\' WM as well as their thinking dispositions or mental styles. WM is thus a defining feature of analytical System 2 processing. System 2 processing is required to solve complex thinking and reasoning problems, although this is not a sufficient condition for valid responses. Most dual-processing theories assume System 1 processing yields intuitive responses that subsequent System 2 deliberation may or may not modify. Stanovich et al. ([@B69]) claim that deliberative reasoning requires overriding System 1. Overriding System 1 and activating System 2 demands an individual\'s executive control, as well as a propensity to think actively and resist the premature closing of problems. Executive control processes thus play a crucial role in analytical System 2 processes (see De Neys and Glumicic, [@B18]; Evans, [@B23]; Thompson, [@B74]). System 2 function relates to general measures of cognitive ability such as IQ (Intelligence Quotient), whereas System 1 function does not (Stanovich, [@B70]). Dual process theories of thinking have been mainly developed in deductive reasoning tasks and there is a lot of evidence for the relevance of both systems of thinking in propositional and syllogistic reasoning (see De Neys, [@B19]; Evans, [@B22]; Barrouillet, [@B5]).
Reasoning is a kind of thinking activity that has a precise starting point, a set of premises, and the goal of drawing a conclusion. In induction, the conclusion involves an increment in semantic information; whereas in deduction the conclusions do not involve any increase in semantic information (Johnson-Laird, [@B38]). Abstract reasoning, even the most elementary kind, is hence a complex phenomenon that requires that individuals follow a sequential process that includes various steps and tasks, and the passage from one to another. A second source of complexity comes from the need to temporarily store and update in WM the diverse representations needed to carry out a reasoning sequence. Indeed, individual differences in reasoning are substantially correlated with the amount of information learners can hold in WM while perform the required inferential reasoning task (Süß et al., [@B64]).
There are diverse theories of deduction, the two most important being that of "mental rules" and "mental models." According to mental rules theories, people possess a set of rules, a sort of "natural logic" from which they reach a conclusion by following a sequence of steps (see Rips, [@B62]; Braine and O\'Brien, [@B8]). The current work, however, has been planned and carried out according to the mental model approach. The mental model theory of reasoning postulates that when individuals face deductive problems they construct models or possibilities of the meaning of assertions consistent with what they describe (see Johnson-Laird, [@B37], [@B39]; Johnson-Laird and Byrne, [@B35]). Mental models only represent what is true according to the premises, but not what is false. For instance, for the conditional "if p, then q" people might construct the following true possibilities: "p and q," "not-p and q," "not-p and not-q." A main assumption of model theory concerns the crucial role of WM in deduction: representing and manipulating models in order to reach a conclusion entails cognitive work and effort. For complex deductive problems, for instance conditionals, several possibilities are true, but people think about as few alternatives as possible because of the limitations of their WM (Johnson-Laird and Byrne, [@B36]). Therefore, reasoners are likely to base most of their inferences from the initial and incomplete representation or models of the premises. The model theory\'s fundamental prediction is thus drawn from the number of models required for reaching a conclusion: the more models, the greater the problem\'s difficulty. When the number of models to be held in mind is reduced, reasoning improves (García-Madruga et al., [@B28]). An inferential conclusion is necessarily valid if it holds in all the models of the premises. Therefore, finding a valid conclusion to complex problems requires that individuals build complete representations of premises and validate initial conclusions by searching for counterexamples that can make them false. From a mental model perspective, García-Madruga et al. ([@B27]) have highlighted the central executive processes as the crucial WM component in the explanation of propositional reasoning performance, as well as its relationship with the two systems of reasoning processes, System 1 and System 2. They found positive correlations between WM and reasoning responses that require high levels of mental word, and negative correlations between WM and reasoning responses that require low levels of mental work. Moreover, high WM individuals were more able to resist superficial responses and rely on the semantic process of constructing models of the meaning of premises from which they drown correct conclusions. Likewise, studies with syllogistic reasoning problems have borne out the crucial role of WM, particularly the executive processes (see Gilhooly et al., [@B31]; Capon et al., [@B12]).
In deductive reasoning, the process of connecting premises to the conclusion is ruled by logic. That is, the conclusions have to be necessary and consistent. The restrictions that rule deductive reasoning are hence metadeductive. Metadeduction involves an individual\'s capacity to reflect on one\'s own logical activity itself and to distinguish between logical validity and reality (see, e.g., Byrne et al., [@B9]). Metadeductive abilities include the implicit understanding of the logical system and its basic concepts of necessity and validity, as well as the explicit use of this knowledge by applying metalogical strategies, such as searching for counterexamples (Moshman, [@B49]). Metacognition has been proposed as an important basis for overrides of System 1 by System 2 (Thompson, [@B74]). In developmental terms, there is a gradual acquisition of metalogical understanding during preadolescent and adolescent years: from 11 to 12 years preadolescents begin to understand the concepts of necessity, consistency and the validity of logical conclusions (Moshman, [@B50]; Santamaría et al., [@B65]). However, complete explicit metalogical capacity only becomes possible in late adolescence and adulthood. In this work, we will use a Deductive Reasoning Test (García-Madruga et al., [@B29]) that assesses both deductive and metadeductive abilities.
Dual process theorists agree that abstract and hypothetical System 2 processes can guide our reasoning on specific tasks toward normative answers, whereas System 1 processes have a mayor influence on everyday judgement. The role of dual processes of thinking in education, and particularly the capacity of the Cognitive Reflection Test (Frederick, [@B25]) to predict academic achievement, have been demonstrated by Gómez-Chacón et al. ([@B32]) within the field of mathematics. According to these authors, the dual processes approach has direct application to mathematical learning. A teacher has to direct attention toward two basic educational objectives: firstly, the promotion of an in-depth understanding of mathematical concepts; and secondly, the inhibition of superficial processes and strategies that otherwise lead to error. Our view is that these two basic objectives are cross-sectional in education, so that they underlie academic achievement not only in mathematics, but across all the diverse school subjects.
The present paper addresses the relationship that abstract reasoning has with academic achievement in secondary school, and focus not only on classic fluid intelligence measures but also on two new measures of cognitive reflection and verbal deduction. We examine the underlying idea that other reasoning measures will be a relevant predictor of academic achievement, even beyond the predictive value of fluid intelligence measures. We argue that abstract reasoning processes are involved in most of the complex learning tasks students commonly face at school, and constitute an important higher-order cognitive ability that underlies academic achievement. In order to extend and clarify this supposition, the main goal of this study was to investigate the relationship between two new reasoning measures, one mathematical (Cognitive Reflection) and the other verbal (Deductive Reasoning), and a third classic visuo-spatial reasoning measure that evaluates fluid intelligence, as well as test their capacity to predict academic achievement in adolescence. In spite of the obvious differences in contents and materials, all of them involve the construction and manipulation of abstract representations in a sequential process that requires supervision and control. Thus, we provide a new estimate of the association between cognitive ability and education by having multiple reasoning tests as predictors of academic achievement at secondary school. Participation of an adolescent sample seemed particularly relevant, as reasoning skills are increasingly important during this developmental period (Barrouillet, [@B5]), when learning activities become more complex. In order to assess this view, we used three diverse kinds of reasoning tests, as we shall describe in more detail in the Methods section: Cognitive Reflection Test (CRT; Frederick, [@B25]), DRT (García-Madruga et al., [@B29]) and Raven\'s Progressive Matrices Test (RPMT; Raven, [@B60], [@B61]; Raven et al., [@B59]). The academic achievement is the criteria for assessing students learning outcomes.
The Cognitive Reflection Test (CRT; Frederick, [@B25]; for a review, see Campitelli and Gerrans, [@B11]; for a metaanalysis, see Brañas-Garza et al., [@B7]) assesses an individual\'s ability to use System 2 processes and resist the tendency to give an immediate response (System 1). Frederick ([@B25]); Toplak et al. ([@B77]) found that many people show a tendency to give the fast incorrect answers that are automatic, superficial, compelling, without thinking deeply enough. Frederick\'s problems are hence difficult and people who give the incorrect responses tend to underestimate their difficulty. Prior research has also evidenced that individuals who perform well tend to perform also well at other general ability tests, and tend to avoid biases in decision-making (Campitelli and Labollita, [@B10]; Toplak et al., [@B76]). Therefore, we expected a high rate of erroneous intuitive responses and a significant positive correlation between intuitive responses and the rating of the difficulty of the problems.
The DRT (García-Madruga et al., [@B29]) assesses both deductive and metadeductive abilities in propositional and syllogistic reasoning. According to mental model theory, most of the deductive problems included in the test (11 out of 15) require the construction of multiple models, that is, require the use of System 2 reflexive thinking. The problems were designed so that they covered a range of difficulty (see Johnson-Laird and Byrne, [@B35], for an analysis of tasks in terms of mental models). We hence expected that most of the problems would be difficult and predicted a close relationship between performance in CRT and DRT.
In order to have a measure of abstract non-verbal reasoning, in this work we used the RPMT (Raven et al., [@B59]) that evaluates fluid intelligence. The RPMT assesses a participant\'s ability to solve new problems by perceiving relationships, inducing rules and completing abstract analogies; in other words, by solving diverse relational reasoning tasks that are carried out in working memory, and governed by the executive processes of "assembly and control" (Carpenter et al., [@B13]). The more difficult problems entail more rules or more difficult rules. Erroneous responses can be yielded by System 1 processes, that is, by a superficial or incomplete kind of reasoning that leads one to give intuitive responses that agree only partially with the rules governing the search for the correct responses. We hence also predict a close relation between performance in CRT and Raven test.
Apart from the general expectations concerning the results in CRT, we expected positive correlations between cognitive reflection, deductive reasoning, and Intelligence measures, given that there are common cognitive, metacognitive and WM\'s executive processes that underlie them. Likewise, we predicted that the three reasoning tasks would correlate with academic achievement. In particular, the correlations between erroneous intuitive responses (System 1) and the ratings of difficulty in CRT with accuracy measures of reasoning (System 2) and academic achievement should be negative. Our second hypothesis includes the main prediction that cognitive reflection, deductive reasoning and intelligence measures should predict a relevant amount of variance of academic achievement. Regarding the CRT, recent findings indicate that it is a predictor of rational thinking performance (Toplak et al., [@B76], [@B77]). Our approach maintains that CRT provides a way to assess a main control executive function: the ability to supervise and inhibit cognitive processes and responses. Therefore, we expected that CRT would be able to predict a relevant amount of variance of deductive inferences, metadeductive inferences and intelligence.
Method {#s2}
======
Participants
------------
Fifty one science students aged between 15.3 and 17.7 years (*M* = 15.97; *SD* = 0.45) participated in the study. They were recruited from a 4th grade (2nd cycle of compulsory secondary education refers to 3rd and 4th grades) of a state secondary school in Fuenlabrada (Madrid, Spain). They were 19 girls and 32 boys (62.35%) distributed in two class-groups (Group A: 27 students, 11 girls; and Group B: 24 students, 8 girls). We chose to select science students because scholastic achievement in sciences usually depends more heavily on reasoning abilities than other subjects like humanities, presumably because those subjects are more hierarchically structured (Deary et al., [@B20]). The school was selected to represent the public educational secondary schools from a medium sized urban area with varied socio-economic classes. Students who did not complete some of the tests and those whose families did not authorize participation in the research were excluded (13) from the final sample (*n* = 51). The sample includes three repeater students that were not excluded from the sample in order to preserving natural class-groups.
Tasks and measures
------------------
### Cognitive reflection test (CRT; Frederick, [@B25])
The CRT assesses an individual\'s disposition and aptitude to be reflective when faced with finding the solution to text-based mathematical reasoning problems. For instance, one of the problems is as follows: "*A bat and a ball cost 1.10 \$ in total. The bat cost 1 \$ more than the ball. How much does the ball cost?\_\_\_\_cents*." The mathematical knowledge required to check the accuracy of the intuitive response is rather simple. However, Frederick ([@B25]); Toplak et al. ([@B77]) found that many people show a tendency to give the fast incorrect answers that are automatic, superficial and compelling ---10 cents---. In other words, wrong answers are typical System 1 responses. The correct answer ---5 cents--- requires System 2 reasoning: only individuals who overcome the System 1 response, deliberately think more in depth and construct a more complete representation can solve the problems. As Campitelli and Gerrans ([@B11]) evidenced, it is a test of cognitive reflection and not just a numeracy test.
A Spanish version of this test was used. Participants are asked to solve three numerical problems. There is no time limit to solve the problems and no alternatives are provided for the participants to choose. Afterwards, participants are also asked to evaluate the percentage of their classmates that will be able to give a correct solution to each problem. The solutions to each problem and the estimated percentages of correct solvers have to be written down in a booklet. We have three kinds of measures the CRT for each problem: correct (System 2) responses, intuitive (System 1) responses, and a rating of the problem\'s difficulty. The total score in CRT is calculated as the number of correct answers. Reliability of Cronbach alpha range between 0.60 and 0.73 values (see, Campitelli and Gerrans, [@B11]).
### Deductive reasoning test (DRT; García-Madruga et al., [@B29])
The DRT is composed of 15 items and covers four types of problems: propositional deductive and propositional metadeductive inferences, and syllogistic deductive and syllogistic metadeductive inferences. All the problems were presented with concrete materials and in a familiar and daily context. The Cronbach\'s alpha in DRT was 0.79.
The propositional deductive inferences were four: two conditional problems, e.g., "If p then q" (Affirmation of Consequent, AC: q then q; and Denying of Antecedent, DA: not-p, then not-q) and two inclusive disjunction problems, "p or q or both" (one affirmative and one negative). In these problems, participants are asked to evaluate the possible conclusions of these four inferences. The problems require the construction of multiple models.
The propositional metadeductive inferences were three truth-table problems in which participants had to analyze the consistency of three problems that consisted of a conditional statement and an assertion. In the first problem, the assertion matched the first initial model of the conditional (p and q) and most of participants should choose the correct response. The second problem required one to construct the second conditional model in which antecedent and consequent are negated (not-p and not-q). Finally, the last problem demanded the construction of the third and most difficult model of conditional in which the antecedent is negated but the consequent is affirmed (not-p and q).
The syllogistic inference task required participants to generate and write the solution to five syllogisms. These syllogisms included the combinations of the four kinds of premises: universal affirmative (A: All X are Y), universal negative (E: No X are Y), particular affirmative (I: Some X are Y) and particular negative (O: Some X are not Y). The syllogisms were of different levels of difficulty. The first syllogism was the easiest: it was the only one-model syllogism. The rest of the syllogisms were multiple-model syllogisms which required the construction of two or three models. Two of the multiple model syllogisms had no valid conclusion (N). Finally, in the syllogistic metadeductive necessity/possibility task, reasoners had to decide if a given conclusion in three syllogistic problems is necessarily true, possible, or impossible. In this case, the first problem was the most difficult.
### Raven\'s test of intelligence (RPMT; Raven et al., [@B59])
The RPMT consists of sixty visual analogy problems. Participants have to identify the relevant features of an array of visual elements and then choose from among eight response alternatives arranged below the matrix which one is the correct element that has to be selected. Each problem consists of a 3 × 3 matrix (9 cells) that contains figural elements, such as geometric figures and lines, in which an element (bottom right position of the matrix) is missing. Participants have to look across the rows and then look down the columns to discover the rules that govern the presentation of the diverse figural elements and then use the rules to determine the missing element. Each problem presents a matrix with a missing element. A Spanish version of this nonverbal test was used. The Cronbach\'s alpha was 0.75.
### Academic achievement
Participants\' academic achievement measures were the students\' achievements of content knowledge learned at the end of school terms; that is, the student\'s final numerical scores obtained in the seven basic subjects that include Language (Spanish), Biology, Physics, History, English, and two subjects of Mathematics. From the specific academic scores, we calculated two overall scores: Overall Math is the mean of the scores in the two mathematics subjects and Overall Achievement is the mean of the other disciplines (without Mathematics). The academic scores were obtained from the school database, but all data are confidential and anonymous. The students\' scores were given by university qualified and professional teachers in the diverse subject matter. The numerical score includes performance on theoretical, practical and attitudinal contents. Teachers evaluate their students using a grading system that ranges from 1 to 10 points in each subject (5 to 10 = approval).
The use of teacher final class grades as summaries of students\' academic achievement poses some difficulties and possible biases concerning the validity and reliability of grades for communicating meaningful information about students\' academic progress (see, Allen, [@B2]). Nevertheless, the assessment of academic performance carried out by teachers in our study was specifically centered on the acquisition of knowledge within each of the different disciplines by giving each student a numerical score. According to Roth et al. ([@B63]), school grades are a good measure of academic achievement since they include information on scholastic performance over a wide period of time, based on different sources (e.g., student\'s motivation) and are less prone to error than other kind of specific achievement tests. Besides, a number of researchers have used academic achievement as measures on academic achievement (e.g., Kuhn and Holling, [@B44]).
Procedure
---------
Participants were tested in one single session lasting 90 min. Each class-group completed the three reasoning tests in their classroom with a short break between them. The order of presentation of the tasks in group A was: RPMT, CRT, and DRT. In the group B the order was the opposite. The three tests were presented to students in specific booklets. Prior to data collection, permission was obtained from parents, teachers and students. Parents were informed that participation in the study involved completing a set of reasoning tasks, which were administered by the researchers in groups. The study and overall procedure were approved by the UNED ethics committee.
Data analyses
-------------
The analyses were performed with the Statistical Package for the Social Sciences (IBM SPSS, 2010). No missing values were registered. Preliminary data screening indicated that the scores on reasoning and achievement measures were reasonably normally distributed. Because the outliers were not extreme, these scores were retained in the analysis. First, the percentage of correct and incorrect responses were calculated. To compare patterns of difficulty in the reasoning tasks, paired-samples *t*-tests were conducted. Second, Pearson bivariate correlations were performed to assess the relationship between cognitive reflection, deductive reasoning, intelligence and academic achievement measures. The strength of the correlations was classified according to the criteria suggested by Cohen ([@B16]): a value of 0.10--0.29 is small; 0.30--0.49 is medium, and 0.50--1.00 is high. Third, to examine the predictive role of variables, the method of standard multiple regression analysis was conducted; that is, all predictor variables were entered in one step. To determine which of the reasoning predictors accounted for most of the variance on the academic achievement scores, we performed two standard multiple regression analyses, one to assess the predictive power of reasoning ability on math achievement, and another one to assess the predictive power of reasoning variables on overall achievement. These analyses included variables that were significantly correlated. In addition, three standard multiple regression analyses were conducted to evaluate how well intelligence, deductive inferences and metadeductive inferences could be predicted from CRT measures. Given that in the CRT correct and intuitive responses practically mirror one another, we introduce in our regression analyses either one or the other, but never both of them.
Results {#s3}
=======
Below are presented the results structured according to the descriptive statistics of reasoning variables and objectives of the present study.
Descriptive statistics and comparisons
--------------------------------------
The results of the diverse measures in the cognitive reflection task (CRT) can be observed in Table [1](#T1){ref-type="table"}. The results of the CRT agreed with our expectations and Frederick\'s ([@B25]) predictions: the CRT was rather difficult, there were more wrong intuitive (58%) than correct responses (33%; *p* \< 0.05, Cohen\'s *d* = 0.64). Participants also clearly underestimated the difficulty of the task. That is, they considered that 76% of their classmates would be able to give correct responses, well above the total percentage of correct answers.
######
The percentages of correct and intuitive responses, and the ratings of difficulty for each of the problems in the cognitive reflection test.
**Problems** **Correct responses (%)** **Intuitive responses (%)** **Difficulty ratings (%)**
-------------------------------- --------------------------- ----------------------------- ----------------------------
1\. The bat and the ball 29 63 84
2\. The machines 41 57 76
3\. The pond and the lily pads 29 57 69
Overall 33 58 76
The results of the DRT can be observed in Table [2](#T2){ref-type="table"}. The percentage of correct responses in multiple model problems was reliably lower than in one model problems (40 and 89%, respectively; *p* \< 0.01, Cohen\'s *d* = 2.23). There was a reliable pattern of decreasing difficulty in the four deductive reasoning tasks: the syllogistic construction task was reliably more difficult than the propositional evaluation task (*p* \< 0.01, Cohen\'s *d* = 0.78), which was more difficult than the propositional truth table task (*p* \< 0.01, Cohen\'s *d* = 0.62), which was later found to be more difficult than the syllogistic necessity/possibility task (*p* \< 0.01, Cohen\'s *d* = 0.37). Likewise, on the whole, deductive inferences were clearly and reliably more difficult than Metadeductive Inferences (*p* \< 0.01, Cohen\'s *d* = 1.51).
######
The percentages of correct responses (Standard deviation) for each of the problems in the Deductive Reasoning test.
**Problems and statements** **Correct responses**
---------------------------------------------- -------------------------------------------- ------------
**PROPOSITIONAL EVALUATION TASK**
Conditional Inferences DA 35% (0.48)
Inclusive Disjunctions AC 25% (0.44)
Aff. 69% (0.47)
Neg. 78% (0.42)
Overall Task 52% (0.27)
**METADEDUCTIVE TRUTH-TABLE TASK**
If p then no-q p q[^\*^](#TN1){ref-type="table-fn"} 94% (0.24)
If p then q not-p q 82% (0.39)
not-p q 24% (0.43)
Overall Task 67% (0.20)
**SYLLOGISTIC CONSTRUCTION TASK**
AI-I[^\*^](#TN1){ref-type="table-fn"} 84% (0.37)
IA-N 2% (0.14)
EI-O 43% (0.50)
EA-O 8% (0.27)
OO-N 31% (0.47)
Overall Task 34% (0.19)
**METADEDUCTIVE NECESSITY/POSSIBILITY TASK**
Possible 45% (0.50)
Possible[^\*^](#TN1){ref-type="table-fn"} 90% (0.30)
Necessary[^\*^](#TN1){ref-type="table-fn"} 88% (0.33)
Overall Task 75% (0.23)
Deductive problems 42% (0.19)
Metadeductive problems 71% (0.19)
Overall correct responses 53% (0.17)
*= one model problem*.
In the RPMT the mean of correct responses was 52.75 and the standard deviation was 3.97 (Max. = 60, Min. = 42). These results show the participants have a high level of fluid intelligence in comparison with their age group (*M* = 48).
Interrelationships between variables
------------------------------------
Table [3](#T3){ref-type="table"} presents the correlations between the mean measures. As can be observed, our predictions included in the first hypothesis were clearly confirmed. The correlations between CRT (correct responses), DRT (deductive and metadeductive inferences) and Intelligence measures were positive and significant. Likewise, accuracy in the three reasoning tasks correlated in reliable way with the two Academic Achievement measures. In DRT, the correlations between multiple and one model problems with Overall Math were: *r* = 0.48 (*p* \< 0.001) and *r* = 0.12 (*p* \> 0.1), respectively; and with Overall Achievement, the correlations were: *r* = 0.46 (*p* \< 0.001) and *r* = 0.14 (*p* \> 0.1), respectively.
######
Pearson correlations between Cognitive Reflection Test (Correct and Intuitive responses, and Ratings of Difficulty), Intelligence (RPMT), Deductive and Metadeductive Inferences, and the two Academic Achievement measures.
***n* = 51** **1** **2** **3** **4** **5** **6** **7** **8**
------------------------------ ------- ------------------------------------------ ------------------------------------------ ------------------------------------------ ------------------------------------------ ------------------------------------------ ------------------------------------------ ------------------------------------------
1\. Reflection Correct 1 −0.94[^\*\*^](#TN3){ref-type="table-fn"} −0.62[^\*\*^](#TN3){ref-type="table-fn"} 0.45[^\*\*^](#TN3){ref-type="table-fn"} 0.46[^\*\*^](#TN3){ref-type="table-fn"} 0.57[^\*\*^](#TN3){ref-type="table-fn"} 0.29[^\*^](#TN2){ref-type="table-fn"} 0.33[^\*\*^](#TN3){ref-type="table-fn"}
2\. Reflection Intuitive 1 0.69[^\*\*^](#TN3){ref-type="table-fn"} −0.47[^\*\*^](#TN3){ref-type="table-fn"} −0.47[^\*\*^](#TN3){ref-type="table-fn"} −0.55[^\*\*^](#TN3){ref-type="table-fn"} −0.35[^\*\*^](#TN3){ref-type="table-fn"} −0.39[^\*\*^](#TN3){ref-type="table-fn"}
3\. Difficulty Ratings 1 −0.33[^\*\*^](#TN3){ref-type="table-fn"} −0.39[^\*\*^](#TN3){ref-type="table-fn"} −0.54[^\*\*^](#TN3){ref-type="table-fn"} −0.37[^\*\*^](#TN3){ref-type="table-fn"} −0.38[^\*\*^](#TN3){ref-type="table-fn"}
4\. Intelligence 1 0.27[^\*^](#TN2){ref-type="table-fn"} 0.33[^\*\*^](#TN3){ref-type="table-fn"} 0.41[^\*\*^](#TN3){ref-type="table-fn"} 0.37[^\*\*^](#TN3){ref-type="table-fn"}
5\. Deductive Inferences 1 0.60[^\*\*^](#TN3){ref-type="table-fn"} 0.48[^\*\*^](#TN3){ref-type="table-fn"} 0.44[^\*\*^](#TN3){ref-type="table-fn"}
6\. Metadeductive Inferences 1 0.29[^\*^](#TN2){ref-type="table-fn"} 0.28[^\*^](#TN2){ref-type="table-fn"}
7\. Overall Mathematics 1 0.89[^\*\*^](#TN3){ref-type="table-fn"}
*M* = 5.58; *SD* = 2.09
8\. Overall Achievement 1
*M* = 6.54; *SD* = 1.57
p \< 0.05;
*p \< 0.01; unidirectional*.
Finally, it may be noted that, as expected, the correlations between erroneous intuitive responses and the ratings of difficulty in CRT with accuracy measures of reasoning and academic achievement were all negative and significant (range between 0.35 and 0.69 values).This finding indicates that the underestimation of difficulty was higher in participants that gave intuitive responses.
Predictive power of variables
-----------------------------
Our second hypothesis claimed the predictive ability of the reasoning measures (Cognitive Reflection---either difficulty ratings or intuitive responses---, Deductive Inferences and Intelligence) on both Academic Achievement scores. Table [4](#T4){ref-type="table"} shows that CRT, DRT and Intelligence measures significantly predicted relevant amounts of variance of the two measures of Academic Achievement: Overall Math and Overall Achievement. The results satisfied our expectations as follows: 33% of the variance of Academic Achievement in Mathematics and 28% of the Academic Achievement in the remaining subjects---Spanish Language, Biology, Physics, History, English---were explained by the abstract reasoning measures. The reliable variables were Deductive inferences and Intelligence in the former, and only Deductive inferences in the overall achievement measure (without Mathematics).
######
Regression analyses of Intelligence, Deductive Inferences and Cognitive Reflection (either Difficulty Ratings or Intuitive Responses) on the two Overall Achievement measures (Overall Math and Overall Achievement without Math).
**Dependent variable** ****R*^2^*** ***F*** ***B*** ***Beta*** ***Sig*.**
------------------------ -------------- -------------------------------------------- --------- ------------ ------------
*Overall Math* 0.33 7.410[^\*\*\*^](#TN5){ref-type="table-fn"}
Intelligence 0.14 0.26 0.04
Deductive Inf. 3.81 0.34 0.01
Diff. Ratings −0.01 −0.14 0.28
*Overall Achievement* 0.28 6.014[^\*\*^](#TN4){ref-type="table-fn"}
Intelligence 0.10 0.22 0.11
Deductive Inf. 3.11 0.32 0.02
Intuit. Resp. −0.60 −0.13 0.39
*p \< 0.01*,
*p \< 0.001; unidirectional*.
Additional multiple regression analyses confirmed that the CRT measures were able to reliably predict relevant amounts of variance of Intelligence (RPMT), deductive and metadeductive inferences (DRT). As confirmed by regression results, Intuitive responses and Difficulty Ratings scores predicted 23% of the variance in Deductive inferences, *F*~(2,\ 48)~ = 0.13, *p* \< 0.01, and fluid Intelligence measures, *F*~(2,\ 48)~ = 6.92, *p* \< 0.01, of which only the Intuitive responses variable (*B* = 0.18, *Beta* = 0.39, *p* \< 0.05; *B* = 4.61, *Beta* = 0.47, *p* = 0.05, respectively) was reliable. The amount of variance of the Metadeductive inferences explained by Correct responses and Difficulty ratings raised to 37%, *F*~(2,\ 48)~ = 11.66, *p* \< 0.001. In this case, although the Difficulty ratings score is close to reaching significance, only the correct responses score was reliable (*B* = 0.19, *Beta* = 0.40, *p* \< 0.05).
Discussion {#s4}
==========
In order to solve new and complex intellectual problems individuals have to think in an active and deep way, using all their WM resources and applying their main executive processes. In this paper, we have investigated abstract reasoning in 15--17 years old students in their final year of Secondary school, using three different reasoning tasks: the CRT, the deductive reasoning task and the fluid intelligence test. Our key predictions concern the relationships between the three kinds of reasoning measures and their capacity to predict academic achievement. With respect to this, we would like to make two main points: first, this study confirms the trend noted in previous research regarding the relationship between reasoning and academic achievement; and second, the present novel findings revealed the relevance of verbal deductive reasoning to predict academic achievement, even beyond a visual fluid intelligence measure.
We provide new evidence on the interplay between the three cognitive variables. As claimed by our first hypothesis, the inter-correlations were all significant and in the predicted direction. Furthermore, on the whole, the magnitude of predicted correlations was medium to high, in particular, the correlations between correct and intuitive responses in the CRT with Intelligence (RPMT), Deductive and Metadeductive inferences (DRT). These results confirm our theoretical conception regarding the tight relationships between the diverse kind of reasoning studied in this paper, as well as the common cognitive, metacognitive, and executive processes that underlie them. An interesting, but not unexpected, result is the high correlation between Cognitive Reflection and Metadeductive Inferences. The greater predictive ability of the CRT on Metadeductive inferences would be explained by the increased involvement of the metacognitive component in the metadeductive tasks. Thus, the pattern of results supports their common metacognitive nature and gives support to the important role that metacognitive processes may play in determining the interventions of System 2, as claimed by Thompson ([@B74]), Thompson et al. ([@B73]). It is interesting to note that, on the whole, the results confirm the relevant role that executive and metacognitive processes play in both verbal and visual abstract reasoning (García-Madruga et al., [@B27]).
The CRT mainly measures the propensity or willingness to be reflective, to think analytically despite having what initially appears to be a suitable response (Pennycook et al., [@B56]). According to our view, this thinking disposition is directly related with executive functioning, since thinking analytically implies the ability to supervise and inhibit cognitive processes and responses. The present findings confirmed that, as expected and found in previous studies (Brañas-Garza et al., [@B7]), most of the participants replied with the wrong intuitive responses in the CRT and also underestimated the difficulty of the task: there was a high correlation between both measures. Only a third of participants were able to resist and override the immediate spontaneous answer and keep thinking to reach the correct one. The analytic thinking carried out by these participants not only allowed them to reach a correct conclusion, it also increased their metacognitive awareness of the difficulty of the problems. Thus, the correlation between correct responses and difficulty ratings measures was highly negative. As explained by Frederick ([@B25]), the correct answers produced by System 2 involve cognitive and motivational effort as well as concentration, and for this reason require more time to solve. These findings contribute to a better understanding of the interaction between System 1 and System 2 at the metacognitive and cognitive levels while performing the problem solving task.
According to our view, this study also allows a more nuanced analysis of the interrelations between variables and particularly the capacity of the CRT to predict verbal relational reasoning (DRT) and visual relational reasoning (RPMT). The proposal that the CRT functions as a measure of analytic System 2 processes supports our hypothesis about its predictive capacity over the DRT and the Raven\'s test. As noted in the Introduction, these two tests assess verbal and visual abstract reasoning, respectively, and require thinking in a reflective, controlled and supervised way. The study provides new novel findings regarding the positive relation of cognitive reflection with respect to reasoning. It also extend a number of previous findings that showed that there is a moderate correlation between CRT and other intelligence measures, such as Wonderlic IQ test (Frederick, [@B25]; Toplak et al., [@B76]).
The results found in DRT show a pattern of increased difficulty, the most difficult of which was the syllogistic deductive task and the easiest the syllogistic metadeductive task. This pattern of results can be explained with reference to two main variables. The first variable is the kind of task participants are required to complete. This involves two sources of difficulty: whether the task is deductive or an easier metadeductive task; and whether the task is a generation or an easier evaluation task. The second variable is the number of models required to solve each problem, as predicted by mental models theory (Johnson-Laird and Byrne, [@B36]): the present results confirm the well-established findings indicating that inferences that call for multiple models are harder than those that call for a single model.
We also investigated the relation between the three reasoning measures and academic achievement measured by means of teacher grades. The inter-correlations between reasoning measures and academic achievement were again in the predicted direction, reliable and moderate. These correlations are somehow lower than the correlations usually found between intelligence and standardized test of achievement (see, for instance, Deary et al., [@B20]; Hannon, [@B34]), but they are quite similar to the correlations found by Soares et al. ([@B68]) by using teachers′ assessments of academic performance and the Reasoning Tests Battery (abstract, verbal, numerical, mechanical and spatial reasoning subtests). Likewise, the findings are informative because they are consistent with those of previous studies (e.g., Frederick, [@B25]; Obrecht et al., [@B54]) that showed moderate and positive correlations between CRT and SAT (Scholastic Assessment Test) scores, one of the most used measures of academic achievement. The results also extend those of Gómez-Chacón et al. ([@B32]) showing the interaction between cognitive reflection, WM, reasoning and mathematical achievement (i.e., academic qualifications) in secondary school. However, the highest correlations found in the present study were those between Academic Achievement and Deductive Inferences. This result is particularly relevant, especially if we notice that it is higher than the correlation found with fluid intelligence as measured by the RPMT.
The second main aim concerns the predictive ability of the reasoning measures on academic achievement. In that respect, we found that the reliable variables were intelligence and deductive reasoning for math achievement, whereas only deductive reasoning was for global academic achievement. The findings are in accordance with those of previous studies that using different methodological approaches evidenced that intelligence is an important predictor of achievement and rate of learning (e.g., Primi et al., [@B57]; Roth et al., [@B63]), either when intelligence is measured by Wechsler Intelligent Scales for Children (Miranda et al., [@B48]) or Kaufman Test of Cognitive Ability (Kaufman et al., [@B41]; Scheiber, [@B66]). Likewise, our results corroborate findings that showed that individual differences in fluid intelligence are strongly related to math achievement (e.g., Primi et al., [@B58]), confirming thus the high ability of intelligence to predict achievement, particularly mathematics achievement (Primi et al., [@B58]) over achievement in more language based subjects (Deary et al., [@B20]). Regarding the predictive effect of reasoning, our results stress the relevance of deductive reasoning as predictor of academic outcomes, in line with previous results. For instance, Bhat ([@B6]) reported that, out of six dimensions of reasoning (i.e., inductive, deductive, linear, conditional, cause-and-effect and analogical reasoning components) assessed on 10th grade students, the maximum involvement was reflected by deductive reasoning followed by cause-effect reasoning and inductive reasoning.
Nevertheless, the most interesting and new finding is that a measure of verbal deductive reasoning predicts Academic Achievement better, even in mathematics, than a visual fluid intelligence measure. As a matter of fact, the main source of novelty is the use itself of an abstract verbal measure of reasoning to predict academic achievement. Oral and written verbal communication is clearly the main way teachers and learners interact and the main source of knowledge acquisition at school. Furthermore, mathematical reasoning and problem solving has a verbal component (see Kintsch and Greeno, [@B42]), and different studies have confirmed the predictive capacity of reading abilities on mathematics (see García-Madruga et al., [@B30]). Our deductive problems involve not only reading comprehension but also, as Raven problems, abstract relational reasoning. Therefore, these results are not entirely unexpected and should not surprise us. The acquisition of new and complex knowledge by means of declarative learning across the various subjects at school relies on basic oral and written communicative abilities, and visual and verbal abstract reasoning.
There might be many factors that have an influence and can also explain the variation in academic achievement. Thus, even though the prediction of academic performance has been widely explored and well-established in relation to cognitive factors such as intelligence or basic cognitive processes, there is a broad consensus that multiple cognitive, personality and motivational variables contribute in an interrelated form to predict individual differences in academic achievement (see, e.g., Ackerman and Heggestad, [@B1]; Chamorro-Premuzic and Furnham, [@B15]). For instance, personality predicts academic performance, even when intelligence and cognitive factors are controlled in adolescents (Noftle and Robins, [@B52]; Leeson et al., [@B47]); academic motivation (Lee and Shute, [@B46]), attitudes (Vilia et al., [@B78]), approaches to learning, prior knowledge, study time, homework parents ′education level are also associated to academic achievement (Núñez et al., [@B53]). Therefore, student related factors, school environment and peer influences not studied here should also be included to better assess students learning in future research.
The current study has some limitations that should be addressed in future studies. In that respect, we should mention the small sample size. The generalization of our results to the wider population of 15--17 years old adolescents pose some difficulties, due not only to the scarce number of participants but also to the special characteristics of our sample: all of them were science students. Hence, our initial results are thus needed for further confirmation by means of studies using a wider sample of students. Reasoning ability appears to be important for predicting scholastic achievement in sciences, whereas other subjects such as languages seem to be more affected by gender-related attributes. Because our study does not seek to examine gender effects and due our preliminary analyses of data showed no gender differences in DRT and academic achievement measures, we did not consider the effect of such variable in the entire sample of the current study. However, gender differences in scholastic achievement as mediated by reasoning ability should be considered as an improvement to be undertaken in future research (see Kuhn and Holling, [@B44]).
Educational implications and conclusion
---------------------------------------
Our research has some evident applications to educational practice. First, it stresses that the evaluation of students\' cognitive abilities at school should not only be based on standard intelligence tests. In this line, Lee and Shute ([@B46]) and Soares et al. ([@B68]) have emphasized the relevance of a multidimensional assessment of students\' cognitive abilities. Without undermining the basic role of intelligence tests at school, our results confirm this approach that highlights that verbal deductive and cognitive reflection abilities might be as important at school as standard intelligence tests. As Toplak et al. ([@B76]) argued, the CRT captures important characteristics of rational thinking that are not measured in other intelligence tests. Both reasoning tasks, CRT and DRT, may be considered by teachers and educational psychologists as a relevant source of information to infer about subsequent school success/failure or to diagnose learning difficulties. Second, our research confirms the importance of a dual-processes approach in education. CRT provides a measure of student propensity to be reflexive and resist intuitive responses when faced with solving quite difficult, even if apparently easy, intellectual problems. As all good teachers are aware, a deeper understanding of concepts and tasks, as well as the inhibition of superficial processes and responses, are crucial in education. CRT allows us to know which students are probably going to need specific attention to grow more reflective and think in a deeper way in their leaning. A third educational corollary of our research is the relationship between metacognitive factors and academic achievement. Two of our measures, CRT and DRT, include an assessment of participants\' metacognitive abilities: the ratings of difficulty in CRT and the metacognitive inferences in DRT. The correlations of these measures with both academic measures confirm the important role of metacognitive issues in education and underline again the required introduction of metacognition in ordinary teaching and learning at school. Finally, the study of individual differences in reasoning measured by the three reasoning tasks is useful to understand better how instruction can be made more effective for more learners. Attempts are being made to elucidate the role of reasoning skills so that appropriate interventions to foster students′ reasoning abilities and preventive strategies can be identified (e.g., Ariës et al., [@B3]). From the perspective of dual processes in reasoning, the fact that System 2 entails conscious reasoning makes it susceptible to educational intervention. The promotion of in-depth understanding and the inhibition of superficial processes and strategies that otherwise lead to error would be considered as relevant instructional strategies.
To end, complex declarative --meaningful- learning at school is a kind of human thinking activity that involves the building of mental representations by integrating external and previously stored information, and their manipulation in WM. Along with other higher-order thinking abilities, meaningful learning is a complex and sequential task that demand learners to activate and use executive functions. In this paper we have presented some preliminary evidences confirming that the presumed outcome of meaningful learning (i.e., academic achievement) in diverse subjects is tightly related to three kinds of abstract reasoning: cognitive reflection, verbal deduction and intelligence. Cognitive reflection provides a measure of executive functioning and therefore underlies reasoners\' performance in verbal and visuo-spatial reasoning tasks. Likewise, the two new measures of abstract reasoning and the classic measure of intelligence are able to predict a substantial amount of variance across the main academic performance measures. Finally, our results confirm the importance of dual processes, verbal deductive and metacognitive approaches in ordinary teaching and learning at school. By this means, the study will provide valuable information about the possible evaluation and intervention by educational psychologist.
Ethics statement {#s5}
================
This study was carried out in accordance with the recommendations of UNED ethics committee with written informed consent from legal tutors of all subjects. All legal tutors gave written informed consent in accordance with the Declaration of Helsinki. The study and overall procedure were approved by the UNED ethics committee.
Author contributions {#s6}
====================
IG-V and JG, Principal Investigators of the R&D Project, substantially contributed to the writing of the manuscript, which was revised critically by JV and GD. All authors contributed to all steps and approved the final version of the manuscript. The authors agree to be accountable for all aspects of the study in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Conflict of interest statement
------------------------------
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank students who voluntarily participated in this research as well as the Spanish state secondary school Instituto La Serna of Fuenlabrada (Comunidad de Madrid).
**Funding.** This research was conducted with the financial support of the research project: EDU2014-56423-R, from the Ministry of Economy and Competitiveness of Spain.
[^1]: Edited by: Jesus de la Fuente, University of Almería, Spain
[^2]: Reviewed by: Adelinda Araujo Candeias, Universidade de Évora, Portugal; Ana Miranda, Universitat de València, Spain
[^3]: This article was submitted to Educational Psychology, a section of the journal Frontiers in Psychology
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Metformin is one of the most widely used glucose-lowering agents to treat type 2 diabetes \[[@CR1]\] and is now recommended as the first-line drug therapy by a recent joint consensus statement \[[@CR2]\]. This unique survivor of the biguanides family is prescribed for its effective anti-hyperglycaemic action, achieving a potent reduction of hepatic glucose production through inhibition of gluconeogenesis \[[@CR3]\]. An increase in peripheral glucose uptake by the drug has also been reported, although the extent and significance of this effect in humans is still being discussed \[[@CR3]\]. Metformin also exerts beneficial effects on circulating lipids and exhibits cardio-protective features in obese patients treated with the drug compared with conventional hypoglycaemic agents \[[@CR4]\]. More recently, epidemiological studies have shown a decrease in cancer incidence in metformin-treated patients, suggesting a new potential application of the drug as an anti-cancer agent \[[@CR5]\]. However, although prescribed since the end of the 1950s and despite extensive effort during the last years, the precise molecular (or biochemical) mechanism or mechanisms of action of metformin remain as yet incompletely understood.
At the beginning of the millennium, a remarkable study by Zhou et al. provided a new mechanistic insight by showing for the first time that metformin activates the AMP-activated protein kinase (AMPK) in hepatocytes; however, they did not provide an underlying mechanism for this activation \[[@CR6]\]. More recently, Shaw et al. reported that activation of the hepatic LKB1/AMPK axis in mice could be involved in the inhibitory effect of metformin on gluconeogenesis, highlighting the putative role of these kinases in the therapeutic action of the drug \[[@CR7]\]. AMPK is a well conserved serine/threonine-protein kinase, which acts as a cellular energy and nutrient sensor, and plays a crucial role in the regulation of metabolic pathways \[[@CR8], [@CR9]\]. AMPK consists of a heterotrimeric complex containing a catalytic subunit α and two regulatory β and γ subunits. Each subunit has several isoforms (α1, α2, β1, β2, γ1, γ2, γ3), which are encoded by distinct genes, giving multiple combinations of holoenzyme with different tissue distribution and cellular localisation \[[@CR8], [@CR9]\]. The α subunit contains a threonine residue (Thr 172) whose phosphorylation by upstream kinases (AMPK kinase \[AMPKK\]), such as the serine/threonine-protein kinase LKB1 or calmodulin-dependent protein kinase kinase IIβ (CaMKKIIβ), is sufficient and necessary for AMPK activation to occur. The β subunit acts as a scaffold to which the two other subunits are bound, and also allows AMPK to sense energy reserves in the form of glycogen \[[@CR8], [@CR9]\]. Binding of AMP to the γ subunit activates AMPK via a complex mechanism involving direct allosteric activation, phosphorylation on Thr172 by AMPKK and inhibition of dephosphorylation of this residue by specific protein phosphatases that remain to be identified \[[@CR8], [@CR9]\]. Thus, any increase in the intracellular AMP:ATP ratio activates AMPK, which results, through phosphorylation of various downstream targets, in concomitant inhibition of energy-consuming processes and stimulation of ATP-generating pathways in order to restore energy balance \[[@CR8], [@CR9]\]. Finally, it has recently been shown that ADP, and therefore the ADP:ATP ratio, could also play a regulatory role in AMPK activities by binding to selective Bateman domains on the γ subunit \[[@CR10]\].
While metformin has been previously shown to decrease cellular ATP levels in hepatocytes owing to its specific inhibition of the mitochondrial respiratory-chain complex 1 \[[@CR11]--[@CR13]\], some of the early works suggested that the drug activates AMPK without increasing the AMP:ATP ratio \[[@CR14], [@CR15]\]. Although these initial conclusions were later refuted by the authors \[[@CR16]\], the absence of clear data linking the mitochondrial effect of metformin to activation of hepatic AMPK still nurtures confusion. Thus, an alternative AMP-independent hypothesis involving mitochondrial-derived peroxynitrite was recently proposed, prolonging the controversy \[[@CR17]\]. Taken together, a clarification of the mechanism by which metformin activates hepatic AMPK is required, especially in humans.
The aim of this study was to test the hypothesis that activation of AMPK by metformin in primary hepatocytes is consecutive to change in cellular energy status owing to metformin's mitochondrial inhibitory effect on the respiratory-chain complex 1.
Methods {#Sec2}
=======
**Materials** All chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA).
**Ethics** All experiments performed in animals were done in accordance with the Institute for Laboratory Animal Research Guide for the Care and Use of Laboratory Animals. All procedures on human tissues have received approval from the university and hospital Ethical Review Boards (St Luc Hospital, Brussels, Belgium).
**Generation of liver-specific Ampkα1/2 knockout mice** The generation of liver-specific *Ampkα1/2* (also known as *Prkaa1/2*)^*−/−*^ mice has been described previously \[[@CR18]\].
**Isolation and primary culture of murine and human hepatocytes** For rodent experiments, liver cells were prepared by the collagenase method of Berry and Friend \[[@CR19]\], modified by Groen et al. \[[@CR20]\], from male Wistar rats (200--300 g) or from male mice (25--30 g) after anaesthesia with sodium pentobarbital (6 mg/100 g body weight) or ketamin/xylazin (8/1 mg/100 g body weight), respectively. For human experiments, hepatocytes were isolated from whole livers or liver segments not used for transplantation, using collagenase P (Roche, Mijdrecht, the Netherlands). For primary culture, rat or human hepatocytes were first seeded for 3 to 4 h on type I collagen-coated dishes (2 × 10^4^ cells/cm^2^) and then cultured in M199 medium (Invitrogen, Leek, the Netherlands) supplemented with antibiotics in the presence of the indicated concentrations of metformin.
**Western blot analysis** Hepatocytes or liver samples were lysed in ice-cold buffer containing: 50 mmol/l HEPES (pH 7.6), 50 mmol/l NaF, 50 mmol/l KCl, 5 mmol/l NaPPi, 1 mmol/l EDTA, 1 mmol/l EGTA, 1 mmol/l dithiothreitol, 5 mmol/l β-glycerophosphate, 1 mmol/l sodium vanadate, 1% NP40 (vol./vol.) and protease inhibitors cocktail (Complete; Roche). Homogenates were centrifuged (16,000 *g*; 15 min, 4°C) and the protein content of the supernatant fraction was determined using a kit (BCA Protein Assay Kit; Pierce, Rockford, IL, USA). Proteins (10--50 μg) were separated by 7--10% SDS-PAGE, followed by transfer to a polyvinylidene fluoride transfer membrane. Membranes were blocked for 1 h at room temperature in TRIS-buffered saline Tween-20 buffer with 5% non-fat dry milk, followed by overnight incubation with phospho-specific or total antibodies (all of them cross-reacting with human, rat and mouse forms, see electronic supplementary material \[ESM\] Table [1](#MOESM1){ref-type=""}). Blots were then incubated for 1 h with horseradish peroxidase-conjugated secondary antibodies at room temperature. Bands were visualised by enhanced chemiluminescence and quantified using Image J (NIH, Bethesda, MD, USA).
**AMPK assays** AMPK activity was assayed either after precipitation with 10% (wt/vol.) polyethylene glycol 6000 or after immunoprecipitation with specific antibodies directed against α1- or α2-AMPK catalytic subunits (Kinasource, Dundee, UK), as described \[[@CR18]\].
**Determination of adenine nucleotide concentrations** Samples of the cell suspension or cultured hepatocytes were lysed in ice-cold HClO~4~-EDTA (5% wt/vol., 25 mmol/l) and centrifuged (13000 *g*, 2 min). The supernatant fractions were immediately neutralised and determination of adenine nucleotides was performed by high-performance liquid chromatography, as previously described \[[@CR18]\].
**RNA purification and quantitative reverse transcription-coupled real-time PCR** RNA was extracted from isolated hepatocytes using an RNA isolation reagent (Tripure; Roche). Total RNA (2 μg) was reverse-transcribed and quantitative real-time PCR then performed with a kit (SYBR Green Core; Bio-Rad, Veenendaal, the Netherlands) on a thermal cycler (MyIQ; Bio-Rad). mRNA expression was normalised to *GAPDH* mRNA content and expressed as arbitrary units. All the primer sets used were designed to span an exon (avoiding eventual amplification of gDNA) and have an efficiency of ∼100 ± 5% (ESM Table [2](#MOESM2){ref-type=""}).
**Determination of mitochondrial oxygen consumption rate in intact and permeabilised hepatocytes** Mouse or human hepatocytes (7--8 mg dry cells per ml) were incubated in a shaking water bath at 37°C in closed vials containing 2 ml Krebs--Ringer bicarbonate-calcium buffer (120 mmol/l NaCl, 4.8 mmol/l KCl, 1.2 mmol/l KH~2~PO~4~, 1.2 mmol/l MgSO~4~, 24 mmol/l NaHCO~3~, 1.3 mmol/l CaCl~2~, pH 7.4) in equilibrium with a gas phase containing O~2~/CO~2~ (19:1) and supplemented with lactate/pyruvate/octanoate (20/2/4 mmol/l) in the presence or not of 5 mmol/l metformin. After 30 min, the cell suspension was saturated again with O~2~/CO~2~ for 1 min and immediately transferred into a stirred oxygraph chamber equipped with a Clark oxygen electrode (HEITO, Paris, France). The mitochondrial oxygen consumption rate (*J*O~2~) was measured at 37°C before and after successive addition of 0.5 μmol/l oligomycin and 150 μmol/l 2,4-dinitrophenol (DNP). To permeabilise hepatocytes, intact cells were first incubated for 30 min as described above, then collected by centrifugation and resuspended in KCl medium (125 mmol/l KCl, 20 mmol/l TRIS-HCl, 1 mmol/l EGTA and 5 mmol/l Pi-TRIS, pH 7.2) containing 200 μg/ml digitonin. After 3 min at 37°C, the permeabilised hepatocytes were transferred to the oxygraph. As indicated, 5 mmol/l glutamate-TRIS plus 2.5 mmol/l malate-TRIS, or 5 mmol/l succinate-TRIS plus 0.5 mmol/l malate-TRIS plus 1.25 μmol/l rotenone were added. *J*O~2~ was measured before and after the successive addition of 1 mmol/l ADP-TRIS, 0.5 μg/ml oligomycin and 50 μmol/l DNP. All the results are expressed in percentage of the maximum activity of cytochrome oxidase, after addition of 0.15 μg/ml antimycin and 1 mmol/l *N*,*N*,*N*′,*N*′-tetramethyl-1,4-phenylenediamine (TMPD) plus 5 mmol/l ascorbate to normalise the results according to respective mitochondrial content in wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice.
**Statistics** All data are expressed as mean ± SEM. Statistical analysis was performed using SPSS 17.0 software package for Windows (SPSS, Chicago, IL, USA) with two-tailed unpaired Student's test or one-way/two-way ANOVA, followed by a Tukey's post hoc test for multiple comparisons. Differences between groups were considered statistically significant at *p* \< 0.05.
Results {#Sec3}
=======
**Effects of metformin on AMPK activity and cellular energy status in rat hepatocytes** In a first set of experiments, the effects of metformin on AMPK activity and AMP:ATP ratio were studied in primary rat hepatocytes. As shown in Fig. [1a](#Fig1){ref-type="fig"}, metformin induced a significant time- and dose-dependent activation of AMPK, which was already detectable at 0.5 mmol/l after 3 h. This dose-dependent effect persisted after 24 h and was paralleled by a concomitant decrease in ATP and increase in AMP intracellular levels (Fig. [1b, c](#Fig1){ref-type="fig"}), resulting in a significant increase of the AMP:ATP ratio (Fig. [1d](#Fig1){ref-type="fig"}).Fig. 1Time- and dose-dependent effects of metformin on AMPK activity and intracellular adenine nucleotide levels in primary rat hepatocytes. Primary cultured rat hepatocytes were incubated for 0, 0.5, 3, 6, 9, 18 and 24 h in M199 medium with different concentrations of metformin (white triangle, 0.5 mmol/l; white square, 1 mmol/l; white diamond, 5 mmol/l) or vehicle (black dots). **a** AMPK activity was measured after polyethylene glycol fractionation. **b** Intracellular ATP and (**c**) AMP concentrations were determined by HPLC, and the AMP:ATP ratios (**d**) calculated. Results are expressed as means ± SEM; *n* = 3--4; \**p* \< 0.05 compared with control-vehicle
**Species-specific activation of AMPK by metformin** We next assessed whether the effects of metformin on AMPK activity and the AMP:ATP ratio were similar in rat and human primary hepatocytes. As shown in Fig. [2](#Fig2){ref-type="fig"}, metformin increased the AMP:ATP ratio, AMPK activity and Ser79-acetyl-CoA carboxylase (ACC) phosphorylation at concentrations higher than 100 μmol/l in rat and human hepatocytes. However, the extent of AMPK activation by metformin was higher in human than in rat hepatocytes at intermediate concentrations of the drug (472% versus 66% at 500 μmol/l, respectively) (Fig. [2a, d](#Fig2){ref-type="fig"}), although the increase in the AMP:ATP ratio was similar (69% versus 54%, respectively) (Fig. [2b, e](#Fig2){ref-type="fig"}), suggesting a species-specific difference in AMPK sensitivity toward AMP, ADP and/or ATP. Strikingly, the determination of isoform-specific AMPKα activities after immunoprecipitation of the α1 or α2 AMPK catalytic subunits revealed that AMPKα2 activity was undetectable in human compared with rat hepatocytes, while AMPKα1 activity was comparable in the basal condition (Fig. [3](#Fig3){ref-type="fig"}). Accordingly, metformin only increased AMPKα1 activity in human hepatocytes, although both AMPKα isoforms were activated by the drug in rat hepatocytes. The apparent species-specific difference in AMPKα2 activity was confirmed by the absence of protein and mRNA expression of this catalytic subunit in human hepatocytes (Fig. [4](#Fig4){ref-type="fig"}). Interestingly, protein levels and mRNA hepatic expression of the β and γ regulatory subunits also significantly differed between species, showing that the β2, γ1 and truncated γ2 forms are expressed in human hepatocytes, whereas the β1, γ1 and γ2 forms are mostly present in rat and mouse hepatocytes (Fig. [4](#Fig4){ref-type="fig"}). We confirmed that these species-specific differences in AMPK isoform levels is also found in whole liver (ESM Fig. [1](#MOESM3){ref-type=""}), excluding experimental bias induced by hepatocytes isolation.Fig. 2Dose-dependent effects of metformin on AMPK activity and the AMP:ATP ratio in primary rat and human hepatocytes. **a**--**c** Primary cultured hepatocytes isolated from rats or (**d--f**) human liver segments were incubated for 24 h in M199 medium with increasing concentrations of metformin (white bars) or vehicle (black bars). **a, d** AMPK activity was measured after polyethylene glycol fractionation and the AMP:ATP ratios (**b**, **e**) were calculated after determination of intracellular adenine nucleotides by HPLC. **c**, **f** The phosphorylation states (p) of Thr172-AMPK and Ser79-ACC, and the abundance of AMPKα were assessed by western blot. Results (**a**, **b**, **d**, **e**) are expressed as means ± SEM; *n* = 3--4; \**p* \< 0.05 compared with control-vehicleFig. 3Effects of metformin on AMPKα1 and α2 activities. **a** Primary cultured hepatocytes isolated from rats or (**b**) human liver segments were incubated for 24 h in M199 medium in the presence of 500 μmol/l metformin (white bars) or vehicle (black bars). AMPKα1 and α2 activity was measured after immunoprecipitation with isoform-specific antibodies. Results are expressed as means ± SEM; *n* = 3--4; \**p* \< 0.05 compared with control-vehicleFig. 4**a** Protein levels of AMPK subunits and (**b**) the corresponding mRNA gene expression were determined in hepatocytes isolated from human (black bars), rat (hatched bars) or mouse (white bars) livers by western blot and RT quantitative PCR, respectively. Western blot analysis (**a**) is representative of three separate experiments and shows that AMPK subunits migrated to their expected molecular mass (α1, 62 kDa; α2, 62 kDa; β1, 38 kDa; β2, 30 kDa; γ1, 38 kDa; γ2~full-length~, ∼75 kDa; γ2~truncated~, ∼55 kDa; γ3, 55 kDa). The western blots for AMPKα2 were performed using three different primary antibodies cross-reacting with human, rat and mouse forms (see ESM Table [1](#MOESM1){ref-type=""}). **b** The results from RT quantitative PCR were corrected for *GAPDH* and are expressed as means ± SEM (*n* = 3). a.u., arbitrary unitsTaken together, the fact that AMPK activation by metformin is associated with an elevated AMP:ATP ratio in hepatocytes strongly suggests that inhibition of the mitochondrial machinery by the drug might be the main underlying mechanism for hepatic activation of the kinase.
**Mechanism(s) of AMPK activation by metformin** To investigate the mechanisms by which metformin affects the cellular energy state, together with the putative involvement of AMPK in this process, we used hepatocytes from wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice. Freshly isolated hepatocytes were incubated with metformin, and AMPK activity and expression, as well as the AMP:ATP ratio and *J*O~2~ were measured. As expected, metformin increased the AMP:ATP ratio, AMPK activity and Ser79-ACC phosphorylation in hepatocytes from wild-type mice (Fig. [5a--c](#Fig5){ref-type="fig"}). In hepatocytes from liver-specific *Ampkα1/2*^*−/−*^ mice, AMPK expression, activity and activation could not be detected (Fig. [5a, b](#Fig5){ref-type="fig"}), but the increase in the AMP:ATP ratio induced by metformin was still present and even significantly higher than in hepatocytes from wild-type mice (Fig. [5c](#Fig5){ref-type="fig"}). Metformin induced a similar inhibition of *J*O~2~ in wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice, an effect that persisted after addition of the mitochondrial oxidative phosphorylation (OXPHOS) uncoupler DNP (Fig. [5d, e](#Fig5){ref-type="fig"}). This clearly indicates that the inhibitory effect of metformin on *J*O~2~ was exerted on the electron transfer chain rather than on downstream step(s) linked to ATP synthesis. The effect of metformin in intact cells from wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice was further investigated after permeabilisation of the plasma membrane by digitonin, allowing the mitochondrial OXPHOS pathway to be investigated in situ. In the presence of glutamate/malate, a substrate for the respiratory-chain complex 1, a significant decrease in mitochondrial respiratory rates could be detected after metformin pre-treatment of cells from wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice, occurring regardless of the mitochondrial energy state (Fig. [5f, g](#Fig5){ref-type="fig"}). By contrast, no differences were observed with succinate/malate, a substrate for the respiratory-chain complex 2 (Fig. [5h, i](#Fig5){ref-type="fig"}). Importantly, a similar specific inhibition of the mitochondrial respiratory-chain complex 1 by metformin was found in freshly isolated human hepatocytes incubated in the same conditions (Fig. [6](#Fig6){ref-type="fig"}).Fig. 5Effects of metformin on AMPK activity, AMP:ATP ratio, and cellular and mitochondrial *J*O~2~ in intact or permeabilised hepatocytes from wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice. Hepatocytes from wild-type *or Ampkα1/2*^−/−^ mice were incubated at 37°C in a Krebs/bicarbonate medium containing lactate/pyruvate/octanoate (20, 2, 4 mmol/l) in the presence of vehicle (black bars) or 5 mmol/l metformin (white bars). After 30 min incubation, AMPK activity in U/g of dry cells (**a**), isoform-specific AMPKα abundance and Ser79-ACC phosphorylation (p) state (**b**), and the AMP:ATP ratio (**c**) were determined. **d**--**i** Oligomycin-sensitive *J*O~2~ was measured after the successive addition of 6 μg/ml oligomycin (basal) and 100 μmol/l DNP in separate experiments, after hepatocytes from wild-type (**d**, **f**, **h**) and liver-specific *Ampkα1/2*^*−/−*^ (**e**, **g**, **i**) mice had been incubated as described above and permeabilised in a KCl medium containing 200 μg/ml digitonine. *J*O~2~ was measured in the presence of glutamate/malate (5 mmol/l; 2.5 mmol/l) (**f**, **g**) or succinate/malate/rotenone (5 mmol/l; 0.5 mmol/l; 1.25 μmol/l) (**h**, **i**) after the successive addition of 1 mmol/l ADP (state 3), 6 μg/ml oligomycin (state 4) and 75 μmol/l DNP. All *J*O~2~ results are expressed as percentage of the maximum activity of cytochrome oxidase (cyt.ox) determined after the final addition of 0.15 μg/ml antimycin and TMPD/ascorbate (1 mmol/l, 5 mmol/l). Results are expressed as means ± SEM; *n* = 3; \**p* \< 0.05 compared with control-vehicle; ^†^*p* \< 0.05 compared with wild-type miceFig. 6Effects of metformin on *J*O~2~ in permeabilised human hepatocytes. Hepatocytes isolated from human liver segments were incubated in the presence of vehicle (black bars) or 5 mmol/l metformin (white bars), and the *J*O~2~ was measured in the presence of (**a**) glutamate/malate (5 mmol/l, 2.5 mmol/l) or (**b**) succinate/malate/rotenone (5 mmol/l; 0.5 mmol/l; 1.25 μmol/l) as described (Fig. [5](#Fig5){ref-type="fig"}). The results are expressed as means ± SEM; *n* = 3; \**p* \< 0.05 compared with control-vehicleCollectively, these results demonstrate that the inhibition of *J*O~2~ and resultant increase in the AMP:ATP ratio induced by metformin in hepatocytes is due to a specific and AMPK-independent inhibition of the mitochondrial respiratory-chain complex 1.
Discussion {#Sec4}
==========
Collectively, our results show that activation of AMPK by metformin in primary hepatocytes from rodents and humans is due to a decrease in cellular energy status resulting from metformin's AMPK-independent and specific inhibition of mitochondrial respiratory-chain complex 1. Strikingly, we report here for the first time that the distribution of AMPK catalytic and regulatory subunits in hepatocytes differs between rodents and humans. This new important finding suggests possible consequences for pharmaceutical strategies focusing on the development of AMPK activators for the treatment of metabolic diseases.
**The mitochondrial respiratory-chain complex 1 is the primary target of metformin** The inhibition of the mitochondrial respiratory-chain complex 1 by metformin was first reported in perfused livers and isolated rat hepatocytes by Leverve's group \[[@CR12]\] and later confirmed in various other cellular models by us \[[@CR21]--[@CR25]\] and others \[[@CR11], [@CR13], [@CR26], [@CR27]\]. Due to its weaker lipophilic property, metformin, in contrast to the other biguanides, induces only mild and specific inhibition of the respiratory-chain complex 1 and does not affect OXPHOS downstream machinery (Figs [5](#Fig5){ref-type="fig"} and [6](#Fig6){ref-type="fig"}; B. Guigas, unpublished results). In the present manuscript, we show for the first time that this mitochondrial effect of metformin is also present in human hepatocytes, confirming that the respiratory-chain complex 1 constitutes the primary target of the drug, whatever the species.Although the exact mechanisms by which metformin decreases complex 1 activity remains unknown to date, it has been shown that this effect requires intact cells \[[@CR12], [@CR22]\] and, at least in hepatocytes, is not prevented by inhibition of nitric oxide synthase or by various reactive oxygen species scavengers \[[@CR12]\]. On the other hand, the mechanism by which metformin activates AMPK is still being discussed. Despite the indisputable mitochondrial effect of the drug, Fryer et al. and Hawley et al. initially reported that metformin activated AMPK in muscle and rat hepatoma cells without affecting the AMP:ATP ratio \[[@CR14], [@CR15]\]. In the present study, we clearly demonstrate that the activation of AMPK by metformin is associated with an increased AMP:ATP ratio resulting from inhibition of the respiratory-chain complex 1. These discrepancies may be explained by technical differences in nucleotide measurement \[[@CR28]\] and subcellular AMP compartmentalisation \[[@CR29]\], or by the fact that immortalised cell lines are highly glycolytic and therefore much less sensitive to impairment of mitochondrial OXPHOS for ATP supply \[[@CR30]\]. Interestingly, metformin was shown to increase cytosolic AMP and AMPK activity in heart \[[@CR29]\]. In addition, the involvement of complex 1 in the activation of AMPK by metformin has also been elegantly demonstrated in pancreatic MIN6 cells, by showing that methyl succinate, a substrate of complex 2 that bypasses the inhibition of complex 1 by the drug, prevented AMPK activation \[[@CR26]\]. Finally, two recent publications also support the finding that metformin activates AMPK via changes in cellular energy status. Thus Hawley et al. showed that the activation of AMPK by metformin is abolished in a cell line stably expressing AMPK complexes that contain an AMP-insensitive γ2 mutant; this indicates that increased cytosolic AMP triggers activation of the kinase by the drug \[[@CR28]\]. On the other hand, Foretz et al. also demonstrated that, contrary to what was initially suggested by others \[[@CR7]\], metformin inhibits hepatic gluconeogenesis by an LKB1- and AMPK-independent mechanism involving an acute decrease in the hepatic energy state \[[@CR31]\].
**AMPK is not required for the inhibitory effect of metformin on respiratory-chain complex 1** In mammalian mitochondria, the respiratory-chain complex 1 is made up of 45 subunits, of which several can be phosphorylated by various Ser/Thr protein kinases, thus affecting respiratory-chain complex 1 activity \[[@CR32]\]. To rule out the possibility that the mitochondrial effect of metformin is secondary to AMPK activation via some putative AMP-independent mechanism, we used AMPK-null hepatocytes isolated from liver-specific *Ampkα1/2*^*−/−*^ mice \[[@CR18], [@CR30]\]. By showing that metformin inhibited the respiratory-chain complex 1 in hepatocytes from wild-type and liver-specific *Ampkα1/2*^*−/−*^ mice, we can clearly rule out the notion that a primary effect of the drug is mediated by AMPK through phosphorylation of some subunits of the respiratory-chain complex 1 and/or other mitochondrial proteins regulating OXPHOS. The more pronounced effect of metformin on the AMP:ATP ratio in primary hepatocytes from liver-specific *Ampkα1/2*^*−/−*^ compared with wild-type mice (Fig. [5](#Fig5){ref-type="fig"}) could be explained by a decrease in ATP synthesis secondary to impaired mitochondrial biogenesis and/or by a lack of AMPK-mediated inhibition of ATP consumption, as previously reported \[[@CR18], [@CR30]\].
**Species-specific differences in hepatic abundance of AMPK subunits** The AMPK protein is a heterotrimer composed of a catalytic α subunit and two regulatory β and γ subunits, in a ratio of 1α:1β:1γ, all of which are required for the formation of a stable and fully functional AMPK complex \[[@CR8], [@CR9]\]. While some minor differences could be shown, depending on the fibre types, for the γ regulatory subunit \[[@CR33]\], it is well established that the α1 and α2 catalytic subunits are both expressed in human and rodent skeletal muscle, and that the α2β2γ1 complex constitutes the majority of AMPK heterotrimers in this metabolic tissue, showing a high degree of consistency between species \[[@CR33]--[@CR37]\]. In the present study, the mRNA expression pattern and protein levels of the AMPK subunits found in rodent livers is in agreement with those previously reported \[[@CR35], [@CR38]\], leading to many possible combinations of heterotrimers containing either α1 or α2 catalytic subunits, and β1, γ1 and γ2 regulatory subunits. As far as we know, our finding showing that the abundance of hepatic AMPK subunits differs in humans compared with rodents is unprecedented. Thus, it might be speculated that the resulting differences in heterotrimeric composition of AMPK could affect the regulation of kinase activity, notably its sensitivity toward AMP. Indeed, it is striking that the activation of AMPK by metformin seemed more potent, although the increase of the AMP:ATP ratio in humans compared with rat primary hepatocytes was similar (Fig. [2](#Fig2){ref-type="fig"}), suggesting that the human AMPK α1β2γ1 or α1β2γ2 truncated complexes could be more sensitive to subtle changes in cellular energy status. However, this is in contradiction with previous in vitro results showing that, in rat liver extract, AMPK complexes containing the α2/β2 isoforms had a greater dependence on AMP \[[@CR35], [@CR39]\]. Interestingly, it has been very recently shown that ADP, like AMP, could also bind to the γ subunit, leading to modulation of the AMP-triggered phosphorylation of AMPK on Thr172 \[[@CR10]\]. It seems therefore possible that subtle species-specific differences in the metformin-induced increase of the ADP:ATP ratio could also be involved. Unfortunately, the determination of intracellular ADP levels by HPLC was technically not possible in our conditions, so further investigations are required to clarify this point. It is worth mentioning that AMPK has also been reported to have a differential and tissue-specific localisation pattern in mammalian cells, with the AMPKα1 subunit being mainly localised in the cytosol, and the AMPKα2 and β2 subunits being localised to the nucleus and cytosolic fractions \[[@CR39]--[@CR42]\]. While not as yet clarified, the subcellular localisation of the kinase might have an important functional role, such as regulation of gene expression by phosphorylation of nuclear targets of AMPK. This would therefore suggest that differences in the composition of AMPK complexes between rodents and human hepatocytes could result in different physiological outcomes, including putative nuclear regulatory functions.Taken together, the species-specific differences shown in AMPKαβγ complexes in the liver imply that pharmaceutical activation of hepatic AMPK could have different effects in rodents and humans.
**Therapeutic relevance of the metformin concentrations used in vitro** As shown in Fig. [1](#Fig1){ref-type="fig"}, the key determinant of AMPK activation by metformin is a balance between the concentration and the time of exposure to the drug. It is worth noting that most of the in vitro experiments reporting activation of AMPK by metformin were generally performed with drug concentrations far above those found in tissues from rodents after oral administration of metformin \[[@CR43], [@CR44]\]. In the present study, we show that hepatic AMPK is activated by metformin at concentrations higher than 100 μmol/l, i.e. about five times the highest plasma level reported in humans after a single oral drug administration \[[@CR45], [@CR46]\]. However, the liver is one of the few organs that can accumulate significant amounts of metformin, with tissue concentrations that reached hundreds of μmol/l in the periportal area \[[@CR43], [@CR44]\]. In addition, the intracellular transport of metformin is mediated by the organic cation transporter 1 (OCT1) \[[@CR47]\], the deletion of which in hepatocytes resulted in impairment of metformin-induced AMPK activation \[[@CR48]\]. Interestingly, increased OCT1 levels and metformin concentrations were recently reported in the liver of mice on a high-fat diet \[[@CR49]\], suggesting that hepatic accumulation of the drug could be even higher in diabetic patients.
Conclusions {#Sec5}
===========
We have demonstrated here that the primary target of metformin in rodent and human hepatocytes is the mitochondrial machinery. The specific inhibition of the respiratory-chain complex 1 by the drug leads to a dose- and time-dependent decrease in ATP levels, which results in a concomitant increase in cytosolic AMP concentrations (and the AMP:ATP ratio), triggering the activation of AMPK. One of the important new findings of this study is the species-specific differences in the profile of AMPK subunits, suggesting that regulation of the kinase, which is well characterised in rodents, could differ significantly in humans. Further investigations are still required, but this point should certainly be considered when developing new pharmacological agents that target AMPK.
Electronic supplementary material {#AppESM1}
---------------------------------
Below is the link to the electronic supplementary material.
######
(PDF 23 kb)
######
(PDF 24 kb)
######
(PDF 112 kb)
The authors are grateful to L. Maisin and M. de Cloedt for technical assistance. This work was supported by FNRS, the Inter University Pole of Attraction, the Association pour l'Etude des Diabètes et des Maladies Métaboliques (ALFEDIAM), Agence Nationale de la Recherche (ANR-10-BLAN-1123-01) and the European Union FP6 programme (Exgenesis/LSHM-CT-2004-005272). B. Guigas was recipient of the ICP "Michel de Visscher" Fellowship.
**Contribution statement** XS, NT and GCvdZ performed experiments, analysed data, and critically reviewed the manuscript; MF and BV provided mouse model and reagents, contributed to discussion and critically reviewed the manuscript; ES and LH contributed to discussion and critically reviewed the manuscript; BG conceptualised the project, performed experiments, analysed data, wrote and edited the manuscript. All authors approved the final version of the manuscript.
**Duality of interest** The authors declare that there is no duality of interest associated with this manuscript.
**Open Access** This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Schizophrenia is a common psychiatric disease with unknown etiology, which symptoms include cognitive dysfunction in addition to positive and negative symptoms \[[@CR1]\]. Cognitive dysfunction, which may reflect genetic susceptibility, is now generally considered to be the core symptom of schizophrenia with its own pathological mechanisms and related to the abnormal development of the nervous system \[[@CR2]\].
The ubiquitin-proteasome system (UPS), which participates in the intracellular signal transduction, synaptic function, plasticity regulation and neurodevelopment, is an important non-lysosomal protein degradation pathway in human cells \[[@CR3], [@CR4]\]. More and more studies show that UPS plays an important role in the pathological process of many neuropsychiatric diseases, including schizophrenia \[[@CR5], [@CR6]\]. The composition of UPS includes ubiquitin (Ub), ubiquitin-activating enzyme (E1), ubiquitin-binding enzyme (E2), ubiquitin ligase (E3) and 26S proteasome, among which E3 is the key factor in the selective degradation mechanism of UPS. The protein encoded by the neural precursor cell-expressed developmentally down-regulated 4 (*NEDD4*) gene belongs to the E3 family of ubiquitin ligase \[[@CR7]\].
Current studies have shown that the *NEDD4* gene is associated with many neuropsychiatric disorders, such as schizophrenia \[[@CR8]\]. Our previous studies have shown that multiple single-nucleotide polymorphisms (SNPs) of the *NEDD4* gene may be associated with schizophrenia, and the TT genotype of rs7162435 locus may be associated with the clinical phenotype of excitement and hostility in patients with schizophrenia \[[@CR9]\]. The protein encoded by the *NEDD4* gene plays an important role in many physiological processes of the central nervous system. Most of these studies focus on degenerative diseases such as Huntington's disease, Parkinson's disease and Alzheimer's disease \[[@CR10], [@CR11]\]. Current research on the *NEDD4* gene and cognitive dysfunction of schizophrenia is rare. Several studies have shown that abnormal expression of the NEDD4 protein in brain affects the ubiquitination of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunit glutamate receptor 1 (GluR1), and the dysfunction of GluR1 may be related to the pathogenesis of schizophrenia and its cognitive dysfunction \[[@CR12]--[@CR14]\].
Therefore, a case-control study was conducted by our research group to investigate the relationship between the *NEDD4* gene polymorphism and cognitive dysfunction in Chinese Han schizophrenia patients.
Methods {#Sec2}
=======
Subject recruitment {#Sec3}
-------------------
The study sample included 296 patients with schizophrenia and 320 healthy physical examinees (Table [1](#Tab1){ref-type="table"}). Table 1The age and gender distributions of patients in case group and control groupGroup*N*Age (χ ± s)GenderMaleFemaleCase group29633.17 ± 10.99108188Control group32032.48 ± 10.73138182t/χ20.0022.825*P*0.4300.093
For the case group, all these patients were hospitalized in the Shandong Mental Health Center from March 2011 to December 2013, most of the patients who come to the hospital can cooperate to complete the psychological test, including MCCB test. The inclusion criteria were as follows: meeting the diagnostic criteria for schizophrenia in the Fourth Edition of the Diagnostic and Statistical Manual for Mental Disorders (DSM-IV); the diagnosis was made by at least two experienced psychiatrists according to the Structured Clinical Interview for DSM-IV (SCID); between the ages of 18 and 65 years old; biology parents were Chinese Han population; education level was junior high school and above; no anti-psychotic drugs were taken at least 1 month before entering the group (the time of inclusion was the day of hospitalization); no modified electric convulsive therapy was performed within 6 months before entering the group. The exclusion criteria were as follows: with other mental disorders except schizophrenia in DSM-IV; with major physical illness; in pregnancy or lactation.
For the control group, a total of 320 healthy physical examinees from March 2011 to December 2013 in Shandong Province were included. The inclusion criteria were as follows: between the ages of 18 and 65 years old; biology parents were Chinese Han population. The exclusion criteria were as follows: with schizophrenia or other mental illnesses in DSM-IV; with family history of mental illness; with major physical illness; in pregnancy or lactation.
Scale selection {#Sec4}
---------------
The general clinical data questionnaire made by our research team was used to collect clinical data of the participants, including their age, gender, education level, marital status, occupation and other disease-related data. The severity of the clinical symptoms of the case group members was assessed by using the Chinese version of Positive and Negative Symptom Scale (PANSS), which has been tested in China for norm and factor analysis and can be used to assess the symptoms of Chinese patients with schizophrenia \[[@CR15]\]. We used the Chinese version of the Matrics Consensus Cognitive Battery (MCCB) to test the cognitive function of case group members. The reliability and validity of the Chinese version had been evaluated, which had been proven to be an effective tool for assessing cognitive deficits of Chinese patients with schizophrenia \[[@CR16]\]. MCCB consists of the following ten subtests: Trail Making Test A (TMT-A); Brief Assessment of Cognition in Schizophrenia (BACS): Symbol Coding; Category fluency Test: Animal Naming (Fluency); Continuous Performance Test-Identical Pairs (CPT-IP); Letter-Number Span Test (LNS); Wechsler Memory Scale-Third Edition (WMS-III): Spatial Span; Hopkins Verbal Learning Test-Revised (HVLT-R); Brief Visuospatial Memory Test-Revised (BVMT-R); Neuropsychological Assessment Battery (NAB): Mazes; Mayer-Salovey-Caruso Emotional Intelligence Test (MSCEIT): Managing Emotions.
SNPs selection {#Sec5}
--------------
The SNPs information of the *NEDD4* gene in Chinese Han population was downloaded from the home page of the the International HapMap Project \[[@CR17]\]. We analyzed these information by using Haploview 4.2 software and selected 9 Tagger SNPs with MAF \> 0.05 and *r*^2^ ≥ 0.8, which were rs2303579, rs3088077, rs7162435, rs11550869, rs16976592, rs9806179, rs2414451, rs12593446 and rs7174459. Functional SNPs were obtained from the National Center for Biotechnology Information (NCBI) Database \[[@CR18]\] and were searched according to the following principles: MAF \> 0.05 in Chinese Han population or located in the 5'UTR region or the 3'UTR region or the exon. The functional SNPs we found were rs2303579, rs2302580, rs62043855, rs3088077, rs7162435, rs11550869, rs3833005, rs16976592, rs9920007, rs17238461 and rs1042477. Based on these information above, rs3088077, rs7162435, rs11550869 in the 3′ UTR region and rs2303579 (chr15:55860531, exon 4), rs62043855 (chr15:55915613, exon 1) in the exon were finally selected for research.
DNA extraction and SNPs genotyping {#Sec6}
----------------------------------
Peripheral venous blood (5 mL) of control group and case group members was collect separately. The blood samples were shaken up in an anticoagulant tube containing 0.5 mol/L EDTA. The anticoagulant tubes containing the blood sample were centrifuged for 10 min at 3000 rpm/min to remove serum and white blood cells. The deoxyribonucleic acid (DNA) of the blood sample was extracted using the modified potassium iodide method. Genotyping was performed using the Taqman probe method of real-time quantitative polymerase chin reaction (Q-PCR). The primers and probes used in the experiment were shown in Table [2](#Tab2){ref-type="table"}. The PCR reaction system was 10 μL, including 0.35 μL each of forward and reverse primers (20 pmol/μL), 0.35 μL each of FAM/HEX probes (10 pmol/μL), 5 μL of Premix, and 1.80 μL of double distilled water and 2 μL of DNA. The reaction conditions of PCR were as follows: denaturation at 95 °C for 3 min; reading at 95 °C for 15 s and 60 °C for 1 min, cycling 40 times; cooling at 4 °C for 30 min; 40 cycles in total. Table 2Primers and probes sequences of SNPsSNPPrimer (5′-3′)Probe (5′-3′)rs3088077Forward: GGCTGTGTTGCTTGATAGATGTTTProbe 1: FAM-TTCCAGACCAcGAGCCCCTAGTG\--TAMRAReverse: GTCCCCAGCTGCAGACCTTProbe 2: HEX-TTCCAGACCAtGAGCCCCTAGTGG-TAMRArs11550869Forward: AAAGCACCTTCTGATTGTATAACACTTTProbe 1: FAM-TCTGGAACTTCTGACAATCTGcCATGA-TAMRAReverse: TACTTAACCTCTCTGGATTCATATTTCTTCProbe 2: HEX-TCTGGAACTTCTGACAATCTGgCATGA-TAMRArs7162435Forward: TACTGCTTTGTGGATCTTTAATGTTTGProbe 1: FAM-ACCAATGGTCAAtAGGATATGCAGGCA-TAMRAReverse: GCAATGGGTAAAAAGTATTAAAGCCTProbe 2: HEX-TGGTCAAcAGGATATGCAGGCAAGA-TAMRArs2303579Forward: TTACTTGACGGTGGAGGTGATGProbe 1: FAM-AAGGCCTGGcTGC\--MGBReverse: TAAGAGAAGATGAAGCCACCATGTAProbe 2: HEX-AAGGCCTGGtTGCT-MGBrs62043855Forward: CAGATGTCCTATGCATGAGCTTAATATProbe 1: FAM-CTGAATCAGAATtAAG-MGBReverse: GTGATTGTAAACCAGAAATGTCAGAAAProbe 2: HEX-CTGAATCAGAATgAAG-MGB
Statistical analysis {#Sec7}
--------------------
In this study, SPSS 21.0 software was used to establish the database, and the goodness-of-fit Chi-square test was used to verify whether the allele and genotype frequencies of the members in the case group and control group corresponded to the Hardy-Weinberg equilibrium (HWE). The chi-square test was used to analyze the differences between the various qualitative data of the study subjects, including gender, family history, education level, allele frequency and genotype frequency, etc. Univariate analysis of variance, analysis of covariance and non-parametric tests were used to compare the differences in cognitive function of patients in the case group, and non-parametric tests were used for those with non-uniform variance. The LSD method was used to perform multiple analyses. All the analyses above were performed using SPSS 21.0 software. We performed a power calculation using the G\*power program based on Cohen's method \[[@CR19]\]. All statistical tests were considered statistically significant at *P* \< 0.05.
Results {#Sec8}
=======
Data from 296 patients with schizophrenia and 320 healthy controls were analyzed. The results showed that there was no statistical difference in gender (χ^2^ = 2.825, *P =* 0.093) and age (t = 0.002, *P =* 0.430) between the two groups (Table [1](#Tab1){ref-type="table"}). The present sample size revealed 99.86% power of detecting a significant association (α \< 0.05) given an effect size index of 0.2 (corresponding to a weak to moderate gene effect).
Hardy-Weinberg equilibrium {#Sec9}
--------------------------
The genotype frequency distributions of the five SNPs in the control group all met HWE. In the case group, the genotype frequency distributions of four SNPs were consistent with HWE, except for rs11550869 (χ^2^ = 4.247, *P* = 0.039), the reasons may be as follows: firstly, the rs11550869 locus may be associated with the occurrence of disease or linked to disease susceptibility genes; secondly, the sample size is not large enough, and it may be necessary to increase the sample size to further clarify this issue.
Comparison of allele and genotype frequencies between case group and control group {#Sec10}
----------------------------------------------------------------------------------
The allele frequencies of rs3088077 (χ^2^ = 15.4464, *P* \< 0.001) and rs2303579 (χ^2^ = 8.301, *P* = 0.004) loci were significantly different between the case group and the control group, while the allele frequencies of rs11550869 (χ^2^ = 1.243, *P* = 0.265), rs7162435 (χ^2^ = 1.706, *P* = 0.192) and rs62043855 (χ^2^ = 1.080, *P* = 0.299) loci were not significantly different (Table [3](#Tab3){ref-type="table"}). Table 3Comparison of allele frequencies of five SNPs in the *NEDD4* gene between case group and control groupSNPAlleleCase groupControl groupχ^2^/*P*rs3088077C354 (0.598)451 (0.705)15.464/\< 0.001\*T238 (0.402)189 (0.295)rs2303579C324 (0.547)402 (0.628)8.301/0.004\*T268 (0.453)238 (0.372)rs11550869G93 (0.157)89 (0.139)1.243/0.265C499 (0.843)571 (0.861)rs7162435T413 (0.678)468 (0.731)1.706/0.192C179 (0.302)172 (0.269)rs62043855T383 (0.647)432 (0.675)1.080/0.299G209 (0.353)208 (0.325)Notes: \* *P* \< 0.05
The genotype frequencies of rs3088077 (χ^2^ = 14.961, *P* = 0.001) and rs2303579 (χ^2^ = 8.106, *P* = 0.017) loci were significantly different between the case group and the control group, while the genotype frequencies of rs11550869 (χ^2^ = 0.730, *P* = 0.694), rs7162435 (χ^2^ = 1.752, *P* = 0.416) and rs62043855 (χ^2^ = 1.047, *P* = 0.593) loci were not significantly different (Table [4](#Tab4){ref-type="table"}). Table 4Comparison of genotype frequencies of five SNPs in the *NEDD4* gene between case group and control groupSNPGenotypeCase groupControl groupχ^2^/*P*rs3088077CC108 (0.365)164 (0.513)14.961/0.001\*TT50 (0.169)33 (0.103)CT138 (0.466)123 (0.384)rs2303579CC95 (0.321)129 (0.403)8.106/0.017\*TT67 (0.226)47 (0.147)CT134 (0.453)144 (0.450)rs11550869GG12 (0.041)10 (0.031)0.730/0.694CC215 (0.726)241 (0.753)GC69 (0.233)69 (0.216)rs7162435TT143 (0.483)171 (0.534)1.752/0.416CC26 (0.088)23 (0.072)TC127 (0.429)126 (0.394)rs62043855TT126 (0.426)148 (0.463)1.047/0.593GG39 (0.132)36 (0.112)TG131 (0.442)136 (0.425)Notes: \* *P* \< 0.05
Comparison of general clinical data and PANSS scores {#Sec11}
----------------------------------------------------
We compared the differences of the general clinical data (age, gender, family history, smoking history, drinking history, history of psychosis, total disease duration) and the PANSS scores (PANSS scale total score, PANSS positive scale score, PANSS negative scale score, and PANSS psychopathology scale score) between different genotypes of each SNP. The results showed that the PANSS psychopathological scores of patients with different genotypes at rs2303579 locus were significantly different (*P* \< 0.05). The ages of patients with different genotypes of rs11550869, rs7162435, and rs62043855 loci were significantly different (*P* \< 0.05), and there was no significant difference in other general clinical data and PANSS scale scores between different genotypes of each five SNPs (*P* \> 0.05).
Comparison of cognitive functions {#Sec12}
---------------------------------
We used one-way analysis of variance to compare the scores of MCCB among the patients with three different genotypes (CC, TT and CT) of rs3088077 locus. Non-parametric tests were used for those with non-uniform variance. After multiple comparison tests, the results showed that there were no significant differences in MCCB subtests scores among patients with CC, TT and CT genotypes at rs3088077 locus (*P* \> 0.05).
Covariance analysis was used to compare the scores of MCCB among the patients with three different genotypes (CC, TT and CT) of rs2303579 locus. The results showed that in the test of WMS-III Spatial Span, the scores of patients with CT genotype were significantly lower than those with CC and TT genotypes (*P* \< 0.05). The scores of patients with TT genotype were significantly higher than those with CT genotype (*P* \< 0.05) in the test of HVLT-R (Table [5](#Tab5){ref-type="table"}). Table 5Comparison of MCCB subtest scores in patients with different genotypes of the rs2303579 locusSubtestrs2303579F*PP*1*P*2*P*3CC (*n* = 95)TT (*n* = 67)CT (*n* = 134)TMT-A48.02 ± 8.59946.52 ± 18.22046.09 ± 11.2570.6410.5270.4310.2760.889BACS: Symbol Coding42.58 ± 10.38942.82 ± 10.02740.78 ± 10.0121.4080.2460.8500.1720.157Fluency49.13 ± 9.90547.75 ± 12.49946.64 ± 11.5831.4140.2450.4670.0940.470CPT-IP42.77 ± 10.61642.58 ± 9.99941.12 ± 11.6570.8450.4310.9470.2430.333LNS46.91 ± 10.07045.54 ± 10.14045.22 ± 12.4200.6230.5370.4270.2850.911WMS-III: Spatial Span45.83 ± 12.23246.16 ± 11.76442.46 ± 13.5072.9170.0560.8430.045\*0.046\*HVLT-R44.20 ± 11.27347.39 ± 10.11343.28 ± 11.3882.7860.0630.0810.5870.020\*BVMT-R43.83 ± 11.79344.22 ± 12.00742.49 ± 11.7640.6140.5420.8320.3690.328NAB: Mazes43.88 ± 9.94743.93 ± 10.45841.82 ± 10.1151.1950.3040.9460.1670.247MSCEIT: Managing Emotions48.33 ± 11.86248.12 ± 12.95747.60 ± 11.5740.0780.9250.8830.6950.846Notes: *P*1 value represented the statistical significance between patients with CC genotype and TT genotype. *P*2 value represented the statistical significance between patients with CC genotype and CT genotype. *P*3 value represented the statistical significance between patients with TT genotype and CT genotype. \* *P* \< 0.05
We used the covariance analysis to compare the scores of MCCB among the three different genotypes of rs11550869 locus, and the same comparison was also performed at rs7162435 and rs62043855 loci. The study of rs62043855 locus showed that patients with TG genotype had significantly lower scores than those with GG genotype (*P* \< 0.05) in the test of NAB Mazes, while no significant difference was found between TG and TT genotype in this subtest (*P* \> 0.05), neither between TT and GG genotype. In the remaining nine subtests of the MCCB, the scores showed no significant difference between patients with different genotype (TG, TT and GG) of this locus (*P* \> 0.05). When it came to rs11550869 and rs7162435 loci, no significant difference between patients with different genotype was found in any subtest of MCCB (*P* \> 0.05) (Table [6](#Tab6){ref-type="table"}). Table 6Comparison of MCCB subtest scores in patients with different genotypes of the rs62043855 locusSubtestrs62043855F*PP*1*P*2*P*3TT(*n* = 126)GG(*n* = 39)TG(*n* = 131)TMT-A47.30 ± 8.98947.49 ± 18.53046.13 ± 13.2060.7060.4940.8220.2430.561BACS: Symbol Coding42.59 ± 10.30143.26 ± 9.33240.66 ± 10.1812.2110.1110.9520.0530.161Fluency48.60 ± 10.77747.54 ± 11.48046.85 ± 11.7661.0140.3640.5160.1580.748CPT-IP42.23 ± 11.05842.49 ± 9.96541.59 ± 11.2150.4000.6710.8620.3790.665LNS46.83 ± 9.85845.28 ± 10.10545.03 ± 12.6341.9620.1420.2300.0580.922WMS-III: Spatial Span45.20 ± 12.49745.54 ± 12.55343.25 ± 13.1720.9490.3880.9080.2150.330HVLT-R44.41 ± 11.46247.03 ± 10.78643.85 ± 10.9411.2460.2890.3050.4530.120BVMT-R44.14 ± 12.06845.21 ± 11.22541.95 ± 11.6711.6690.1900.6210.1440.131NAB: Mazes43.61 ± 10.29945.44 ± 9.21041.60 ± 10.1622.7230.0670.3700.0920.03**\***MSCEIT: Managing Emotions47.72 ± 11.47348.26 ± 14.36548.08 ± 11.7170.0130.9870.8800.9090.942Notes: *P*1 value represented the statistical significance between patients with TT genotype and GG genotype. *P*2 value represented the statistical significance between patients with TT genotype and TG genotype. *P*3 value represented the statistical significance between patients with GG genotype and TG genotype. \* *P* \< 0.05
Discussion {#Sec13}
==========
At present, studies have shown that the *NEDD4* gene may be related to the pathological process of schizophrenia. Primary study of our group showed that several loci of the gene may be associated with the occurrence of schizophrenia and the clinical characteristics of patients, however, there is no study to investigate whether the gene is associated with cognitive impairment of patients with schizophrenia. Therefore, we conducted the above studies to analyze whether there is an association between the *NEDD4* gene and cognitive impairment of patients with schizophrenia. The results of the study were analyzed as follows:
The score of WMS-III Spatial Span in MCCB represents the nonverbal working memory of the testee \[[@CR20]\]. The multicomponent model of working memory proposed by Baddeley et al. believed that working memory should be elaborated into a three component system, including the visuo-spatial scratchpad, the articulatory loop and the central executive system \[[@CR21]\]. At least two types of working memory systems have been discovered, including verbal working memory for processing linguistic information and spatial working memory, which tested by WMS-III Spatial Span in MCCB, for processing spatial information. The spatial working memory plays an important role in the spatial orientation and the solution of visuospatial problems. Glahn found that spatial working memory impairment was one of the neurocognitive deficits that may mark the genetic predisposition to schizophrenia \[[@CR22]\], and his findings suggested that deficits of short-term spatial mnemonic processing may be an effective endophenotypic marker for schizophrenia. Combined with our findings, the scores of patients with CT genotype of rs2303579 locus were significantly lower than those with CC and TT genotypes in the test of WMS-III Spatial Span. So it can be considered that patients with CT genotype of rs2303579 locus has more severe spatial working memory impairments, which means the patients with this genotype have more serious defects in temporary storage and initial processing of visual and spatial information than patients with other two genotypes of rs2303579 locus. It is already known that the mutation of C \> T at rs2303579 locus can cause amino acid residues to change from Serine to Asparagine \[[@CR18]\]. It remains to be further studied whether this amino acid change affects the function of NEDD4 protein, and then leads to the deficiency of spatial working memory in patients with schizophrenia.
The test of HVLT-R in MCCB can be used to test whether verbal memory and learning ability of schizophrenia patients are impaired \[[@CR20]\]. Warrick et al. found that people with high-risk schizophrenia showed impaired verbal memory and learning functions before the full expression of psychotic illness, so verbal memory dysfunction should be a genetic endophenotype of the disease \[[@CR23]\]. However, Gildas et al. found that depression, low processing speed and selective attention deficits may affect the verbal memory function of patients with schizophrenia, they believed that patient's verbal memory dysfunction was caused by the above symptoms rather than a primary feature of the disease \[[@CR24]\]. Therefore, there is still no unified view on whether the verbal memory dysfunction is the endophenotype of patients with schizophrenia. Combining our analysis results, the scores of patients with TT genotype of rs2303579 locus were significantly higher than those with CT genotype (*P* \< 0.05) in the test of HVLT-R. Although there was no significant statistical difference compared with those individuals with CC genotype, the TT genotype individuals also showed an increasing trend (*P* = 0.08), so it can be concluded that patients with TT genotype of rs2303579 locus have better verbal memory than patients with CC and CT genotypes. Therefore, we speculated that rs2303579 locus polymorphism may be associated with verbal memory impairment of patients with schizophrenia. The verbal memory impairment is related to genetic factors and may be one of the endophenotype of this disease.
The cognitive function tested by the test of NAB Mazes in MCCB is the ability of reasoning and problem solving \[[@CR20]\]. The ability of reasoning and problem solving is part of the executive functions, which are very important to the appropriate behavior and interact with other cognitive dysfunctions in schizophrenia patients. The NAB Mazes required the patient to retain the information on the pathways that have been previously passed, and then infer the next step. The assessment found that the schizophrenia patient had a defect in the function of the previous information maintaining, which may be one of the important factors for the deficiency of its executive function \[[@CR25]\]. The NAB Mazes can be used to reflect the cognitive ability of the prefrontal cortex of the patient, including the patient's foresight, planning ability, and impulse control ability \[[@CR26]\]. The cognitive impairment may be involved with a wide area of the prefrontal cortex, and its specific pathological mechanism has not been elucidated \[[@CR27]\]. Combined with our study, patients with TG genotype of rs62043855 locus had a significantly lower score than those with GG genotype (*P* \< 0.05) in the NAB. There was no statistically significant difference between TG and TT genotype individuals, but TG genotype individuals also showed a decreasing trend (*P* = 0.092). Therefore, we believed that the genetic polymorphism of rs62043855 locus was associated with the impairment of the ability to reason and solve the problem in schizophrenia patients, the TG genotype patients' foresight, planning ability, and impulsive control ability were worse than those of other genotypes. As we already know, the mutation of T \> G at rs62043855 locus can cause amino acid residues to change from Asparagine to Histidine \[[@CR18]\], which may cause changes in the function of NEDD4 protein. However, the relationship among the amino acid residues changing, NEDD4 protein and the impairment of the ability to reason and solve the problem in schizophrenia patients still remains to be further studied.
There were still some limitations in our research. First of all, only the MCCB was used to perform cognitive tests on patients, without other cognitive testing tools such as University of California, San Diego performance-based skill assessment (UPSA), so the cognitive deficits of schizophrenia patients may not be fully reflected. Secondly, our study only analyzed the association between a single locus of the *NEDD4* gene and cognitive dysfunction in schizophrenia. However, schizophrenia is a complex disease with multiple genes involved, therefore, the genetic mechanism of cognitive dysfunction in schizophrenia patients needs to be further discussed in terms of gene-gene interaction and gene-environment interaction.
Conclusions {#Sec14}
===========
In our research, the *NEDD4* gene polymorphism of rs3088077 and rs2303579 loci was associated with schizophrenia in Chinese Han population. The rs2303579 locus was associated with the disorders of spatial working memory and verbal memory in schizophrenia, in which the CT genotype was a risk factor of spatial working memory impairment, and the TT genotype was a protective factor of verbal memory. The rs62043855 locus was associated with the disorders of reasoning skills and problem-solving capacity in schizophrenia, in which the TG genotype was the risk factor of these disorders.
AMPA
: Α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid
BACS
: Brief Assessment of Cognition in Schizophrenia
BVMT-R
: Brief Visuospatial Memory Test-Revised
CPT-IP
: Continuous Performance Test-Identical Pairs
DNA
: Deoxyribonucleic acid
DSM-IV
: Fourth Edition of the Diagnostic and Statistical Manual for Mental Disorders
E1
: Ubiquitin-activating enzyme
E2
: Ubiquitin-binding enzyme
E3
: Ubiquitin ligase
GluR1
: Glutamate receptor 1
HVLT-R
: Hopkins Verbal Learning Test-Revised
HWE
: Hardy-Weinberg equilibrium
LNS
: Letter-Number Span Test
MCCB
: Matrics Consensus Cognitive Battery
MSCEIT
: Mayer-Salovey-Caruso Emotional Intelligence Test
NAB
: Neuropsychological Assessment Battery
NCBI
: National Center for Biotechnology Information
*NEDD4*
: Neural precursor cell expressed developmentally down-regulated 4
PANSS
: Positive and Negative Syndrome Scale
Q-PCR
: Quantitative polymerase chin reaction
SCID
: Structured Clinical Interview for DSM-IV
SNP
: Single-nucleotide polymorphism
TMT-A
: Trail Making Test A
Ub
: Ubiquitin
UPS
: Ubiquitin-proteasome system;
UPSA
: University of California, San Diego performance-based skill assessment
WMS-III
: Wechsler Memory Scale-Third Edition;
**Publisher's Note**
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Chao Han and Kaiyan Cui contributed equally to this work.
We thank all participant for their support in this study.
LL designed the research; CH and XB performed the experiments and analyzed the data; KC and XB conducted sample collection and contributed reagents/materials tools; LL, KC, LW, LY, XB and MS were responsible for clinical diagnosis; CH was responsible for manuscript writing. All authors read and approved the final manuscript.
This study was supported by Medicine and Health Science Technology Development Program of Shandong province (2015WSB06012). The funding body played no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.
Researchers interested in the study may contact corresponding author to obtain relevant data via email: liulf521\@163.com.
The study was approved by the Ethics Committee of the Mental Health Center of Shandong Province, reference number: (2015) Ethics Review (R08). All participants signed the informed consent before participating in the study.
Not applicable.
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Contamination of surface and groundwater by trace metals results in the deterioration of water quality which affect human health as well as the health of aquatic ecosystem (Krishna et al. [@CR4]; Bataynen [@CR2]). Trace metals in the aquatic environment creates an immense threat to the existence of organisms thriving in the area and the ecological integrity of the habitat, particularly as trace metals may enter the food chains, persist in the environment, bioaccumulate and bio-magnify. Though some trace metals in lower concentrations play important roles in metabolic processes of living organisms, high concentration have been observed to be toxic for human and aquatic life (Ouyang et al. [@CR4]; Adepoju-Bello et al. [@CR1]). High concentration of trace metals in water sources may lead to adverse effects such as deformities, cancer and bad health of aquatic animals and their terrestrial predators (Coeurdassier et al. [@CR2]; Volpe et al. [@CR15]; Kavacar et al. [@CR11]). In humans trace metals above certain concentrations may lead to health problems including liver diseases, kidney problems and Geno toxic carcinogens (Knight et al. [@CR12]; Gambrell [@CR2]).
Metals enter rivers and lakes through a variety of sources such as eroded minerals within sediments, leaching of ore deposits, decomposing dead organic matter, fallout of atmospheric particulate and volcanism extruded products or anthropogenic sources including the discharge of liquid and solid waste, industrial or domestic effluents, channel and lake dredging etc. (Marcovecchio et al. [@CR11]). Trace metals enter the human body through several routes such as food chain, direct ingestion, dermal contact, fume inhalation and particles through mouth and nose (Li and Zhang [@CR19]; Wu et al. [@CR11]; USEPA [@CR12], [@CR4]).
For effective assessment of water quality it is important to identify potential human health effects of pollution in water. The traditional method for evaluating health effects directly compare the measured values with permissible limits, but it is not sufficiently reliable to provide detailed hazard levels and identify contaminants of the most concern. Health risk assessment is an important tool for estimating the potential health impact in aquatic ecosystems caused by various contaminants (Wu et al. [@CR12]; Iqbal and Shah [@CR2]). This method has been applied to evaluate the potential adverse health effects from exposure to contaminated water (Kavacar et al. [@CR11]; Hartley et al. [@CR9]; Sun et al. [@CR36]). Although ingestion is considered the primary route of exposure to chemical contamination in drinking water sources inhalation and dermal absorption are increasingly being taken into account as important exposure pathways.
Lakes have important multi usage components including source of drinking water, irrigation, shipping, fishing, land scape entertainment and hydro-energy production (Yu et al. [@CR39]). The Bosomtwe Crater lake which is a natural inland freshwater that originated from meteorite impact (Koeberl et al. [@CR9]) and serve many functions including water for drinking and domestic use, fishing, transportation, tourism and landscape entertainment. Therefore safeguarding the quality of water in the lake and its surroundings is a great responsibility of the Government of Ghana, researchers and environmentalist for the conservation of this important water resource and world heritage site. However, there is limited information on the effects of the different anthropogenic activities on the water quality and the resultant health effects of the Bosomtwe Crater Lake. The aim of the study is to determine the levels of the selected metals (As, Cu, Fe, Cr, Cd and Pb) in the lake and groundwater and to evaluate the health risk associated with expose to these metals through oral ingestion and absorption through the skin. These metals were selected based on the dominant anthropogenic activities around the lake which include agriculture, proliferation of artisanal gold mining which contribute to soil pollution and land degradation, with the attendant exposure of the environment to trace metals pollution.
Methods {#Sec2}
=======
The study area {#Sec3}
--------------
Lake Bosomtwe is centered at 06°32′N and 1°25′W and is one of the nineteen (19) confirmed impact structures known on earth and has an age of 1.07 million years. According to Koeberl et al. ([@CR9]) it is associated with one of the four tektites strewn fields in the Cote d'Ivoire tektite field. The lake is completely filled with water in a circular structure of roughly 8.5 km in diameter with a rim-to-rim diameter of about 10.5 km and has a depth of about 78--80 cm in its central part (Watkins [@CR36]). The vegetation of the region around the lake is widely that of a dense tropical rainforest. The hydrogeology of the basin is dominated by aquifers of the crystalline basement rocks and the Birimian Province. Groundwater occurs mainly in the Birimian geological formations made up of the Lower Birimian (metasediment rocks) and the Upper Birimian (metavolcanic rocks). The Lower Birimian comprises of over 80 % of the total landmass of the basin while the Upper Birimian crops out in the eastern and extreme southern sections of the basin. Since the lake is a hydrological basin (Turner et al. [@CR11]) all pollutants remain in the lake resulting in a complex and fragile ecosystem as they do not have self-cleaning ability and therefore readily accumulate pollutants (Lokeshwari and Chamdrappa [@CR15]). The main occupation of the people around the lake is fishing and farming but quite recently there is a proliferation of artisanal gold activities around the area.
Sampling and analysis {#Sec4}
---------------------
Water samples were collected from the lake and bore holes from seven human settlements around the Bosomtwe Crater Lake namely at Anyinatiase, Nkowi, Adwafo, Abrodwum, Obo, Abono and Abease as indicated in Fig. [1](#Fig1){ref-type="fig"}. Sampling was done every 3 months from November, 2012 to August, 2013.Fig. 1Map of study area showing towns where samples were taken in the Bosomtwe Crater Lake area
Samples were collected into sterile screw capped plastic containers which had been washed with detergents, 4.0 mol dm^−3^ nitric acid solution and distilled water and dried in an oven. The pH, conductivity total dissolved solids and temperature were measured on the site using CyberScan PC 650 multimeter whilst turbidity was measured with a Hanna HI 93414 turbidity meter. The samples were then filtered through pre-washed 0.45 μm Millipore nitrocellulose filters to remove any suspensions, acidified with 6 M Analar HNO~3~ (2 ml l^−1^) to keep pH \< 2 (USEPA [@CR39]), transported in an ice chest to the laboratory and stored at −4 °C in a refrigerator. The trace metals were determined using Atomic Absorption Spectrophotometer (Model; Varian 220) using air-acetylene flame at a temperature of about 2300 °C. The appropriate hollow cathode lamps (HCL) for each of the elements As, Fe, Pb, Zn, Cr, Cd and Ni were employed as common radiation source at operating wavelengths of 193.7, 248.3, 217.0, 213.9, 357.9, 228.8 and 232.0 nm respectively.
Quality control {#Sec5}
---------------
Quality control measures for the analysis were implemented through the analytical protocols, including sampling and sample preservation, instrument performance evaluation, calibration of instrument, recovery and reagent blank and replicate analyses. The instrument performance evaluation involved optimizing instrument parameters followed by sensitivity check.
Statistical analysis {#Sec6}
--------------------
The results were analyzed statistically employing Microsoft Excel (2010 edition) and statistical Package for Social Science (IBM SPSS version 20). Multivariate statistics in terms of principal component analysis (PCA) cluster analysis were carried out using the varimax-normalized rotation on the data (Shah et al. [@CR2]; Iqbal and Shah [@CR2]).
Health risk assessment methodologies {#Sec7}
------------------------------------
The human health risk assessment methodologies for aquatic ecosystems has been described literature (Li and Zhang [@CR19]; USEPA [@CR12], [@CR19]; Wu et al. [@CR11]). The ingestion and dermal absorption are common for water exposure (USEPA [@CR12], [@CR19]; Wu et al. [@CR11]). The numeric expressions for risk assessment as obtained from the USEPA Risk Assessment Guidance for Superfund (RAGS) methodology (USEPA [@CR12]) are given as follows:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$D_{\text{ing}} = \frac{Cwater \times IR \times EF \times ED}{BW \times AT}$$\end{document}$$ $$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$D_{\text{derm}} = \frac{Cwater \times SA \times KP \times ET \times EF \times ED \times CF}{BW \times AT}$$\end{document}$$where *D* ~ing~ is exposure dose through ingestion of water (μg/kg/day); *D* ~derm~ is exposure dose through dermal absorption (μg/kg/day); *C* water is concentration of the estimated metals in water (μg/L); IR is ingestion rate (2.2 L/day for adults; 1.8 L/day for children); EF is exposure frequency (350 days/year); ED is exposure duration (70 years for adults; and 6 years for children); BW is average body weight (70 kg for adults; 15 kg for children); AT is averaging time (25,550 days for adults; 2190 days for children); SA is exposed skin area (18,000 cm^2^ for adults; 6600 cm^2^ for children); ET is exposure time (0.58 h/day for adults; 1 h/day for children); CF is unit conversion factor (0.001 L/cm^3^); and *Kp* is dermal permeability coefficient (cm/h). The dermal permeability coefficient for Fe, Pb, Zn, Cr and Ni are given as 1.0 × 10^−3^, 4.0 × 10^−3^, 6.0 × 10^−3^, 2.0 × 10^−3^ and 4.0 × 10^−3^ cm/h respectively (USEPA [@CR12], [@CR19]; Wu et al. [@CR11]; Liang et al. [@CR2]).
Potential non-carcinogenic risks for exposure to contaminants were determined by comparing the calculated contaminant exposures from each exposure route with the reference dose (RfD) (USEPA [@CR12]). The hazard quotient (HQ) which is a numeric estimate of the systemic toxicity potential posed by a single element within a single route of exposure was calculated using the relation:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$HQ_{\text{ing/derm}} = \frac{{D_{ing/derm} }}{{RfD_{ing/derm} }}$$\end{document}$$where HQ~ing/derm~ is hazard quotient via ingestion or dermal contact and Rf*D* ~ing/derm~ is oral/dermal reference dose (μg/kg/day). The Rf*D* ~ing~ and Rf*D* ~derm~ values were obtained from the literature (Li and Zhang [@CR19]; USEPA [@CR12]; Wu et al. [@CR11]; Liang et al. [@CR2]).
The overall potential for non-carcinogenic effects posed by more than one element was evaluated by integrating the computed HQs for each element and expressed as a hazard index (HI) (USEPA [@CR12]):$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\text{HI}} = \mathop \sum \limits_{i = 1}^{n} HQ_{ing/derm}$$\end{document}$$where HI~ing/derm~ is hazard index via ingestion or dermal contact. Chronic daily intake (CDI) was calculated using the relation:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\text{CDI}}_{\text{ing}} = {\text{C}}_{\text{water}} \times \frac{DI}{BW}$$\end{document}$$where C~water~, DI and BW represent the concentration of trace metal in water in (µg/kg), average daily intake of water and body weight respectively.
Cancer risk (CR) was calculated using the formula:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\text{CR}}_{\text{ing}} = \frac{{D_{ing} }}{{SF_{ing} }}$$\end{document}$$where SF~ing~ is the cancer slop factor. The SF~ing~ for Pb is 8.5 and Cr is 5.0 × 10^2^ mg/kg/day (USEPA [@CR12], [@CR19]; Yu et al. [@CR4]; Vieira et al. [@CR2]).
Results and discussion {#Sec8}
======================
Table [1](#Tab1){ref-type="table"} shows the mean pH, temperature and conductivity of water samples from Lake Bosomtwe and Groundwater in some towns around the lake in the wet and dry seasons. Table [2](#Tab2){ref-type="table"} also show the mean concentrations of As, Fe, Pb, Zn, Cr, Cd and Ni in water samples from Lake Bosomtwe and Groundwater in some towns around the lake in the wet and dry seasons. The values show that the water from the lake in the dry season is alkaline with the pH values above the recommended WHO standards of 6.50--8.50. However, the pH for lake water in the wet season and those of the groundwater for both dry and wet seasons were observed to be within the WHO range. These values indicate that there was not much difference in the pH in both the water from the lake and groundwater in the surroundings. Thus there in no significant variability of pH due to seasonal contamination. Almost all the water from the lake has the average temperature of 30 °C and the water from the groundwater has the average temperature of 29 °C. Generally the mean conductivities were high indicating that the water contains large amounts of ions which were responsible for such conductivities (APHA [@CR2]: Jain et al. [@CR15]). The health effects for consuming water with high conductivity are the disturbances of salt and water balance which have adverse effects on some myocardial patients and individuals with high blood pressure (Fatoki and Awofulu [@CR11]). The total dissolved solids which may be a measure of the dissolution mechanism of inorganic and organic materials in water for both groundwater and lake water were generally low and below the WHO value of 1000 mg/L for both seasons. Turbidity was also low for both sources of water for both dry and wet seasons and were also below the WHO limit of 5 NTU. Turbidity in water is caused by colloidal matters or suspended particles that obstruct light transmission through the water may originate from the presence of inorganic or organic matter or the combination of both.Table 1Mean values of some physical parameters in lake and groundwater samples collected from seven human settlements along the Bosomtwe Crater Lake in dry and wet seasonsSeasonSettlementLake waterGroundwaterpHTemperature (°C)TDS (mg/L)Turbidity/NTUConductivity (µs cm^−1^)pHTemperature (°C)TDS (mg/L)Turbidity/NTUConductivity (µs cm^−1)^Dry (\*N = 105)Anyinatiase9.83 ± 0.0130.00 ± 0.00237.40 ± 5.461.32 ± 0.031263.40 ± 8.636.77 ± 0.2729.00 ± 0.00197.99 ± 7.011.00 ± 0.011375.60 ± 177.91Nkowi9.77 ± 0.0130.00 ± 0.00215.25 ± 3.580.71 ± 0.011307.30 ± 73.117.19 ± 0.3031.00 ± 1.41178.79 ± 3.210.97 ± 0.011277.10 ± 3.96Adwafo9.82 ± 0.0230.00 ± 0.00215.80 ± 3.450.87 ± 0.041201.10 ± 41.447.18 ± 0.1429.00 ± 1.41200.46 ± 2.320.80 ± 0.031395.00 ± 26.09Abrodwum9.75 ± 0.0130.00 ± 0.00206.03 ± 4.681.06. ± 0.021250.45 ± 8.277.42 ± 0.1628.00 ± 0.00198.77 ± 4.330.71 ± 0.021242.35 ± 9.69Obo9.71 ± 0.0129.00 ± 0.00197.95 ± 5.210.90 ± 0.021787.05 ± 172.757.14 ± 0.0430.00 ± 0.00189.33 ± 1.240.67 ± 0.011285.28 ± 0.95Abono8.98 ± 0.0028.50 ± 0.00208.10 ± 4.280.88. ± 0.041266.23 ± 6.757.19 ± 0.0930.00 ± 0.00199.87 ± 2.450.42 ± 0.011264.60 ± 5.37Abease9.82 ± 0.0230.00 ± 0.00197.95 ± 3.321.00 ± 0.021197.35 ± 60.607.34 ± 0.2329.50 ± 0.71200.06 ± 3.420.50 ± 0.001238.63 ± 24.71Wet (\*N = 105)Anyinatiase6.88 ± 0.1129.00 ± 0.00206.70 ± 4.110.90 ± 0.021241.85 ± 6.726.77 ± 0.2729.00 ± 0.00256.87 ± 5.580.90. ± 0.021356.1 ± 150.40Nkowi6.99 ± 0.0131.50 ± 0.71199 67 ± 3.241.64 ± 0.011264.90 ± 37.907.19 ± 0.3031.00 ± 1.41211.43 ± 3.421.10 ± 0.031277.05 ± 3.89Adwafo7.14 ± 0.0829.50 ± 0.71180.14 ± 4.380.95 ± 0.031212.65 ± 24.827.18 ± 0.1429.00 ± 1.41209.76 ± 5.620.90 ± 0.011392.38 ± 24.25Abrodwum7.48 ± 0.0828.00 ± 0.00200.84 ± 3.681.23 ± 0.031260.00 ± 10.897.42 ± 0.1628.00 ± 0.00198.77 ± 4.180.81 ± 0.011218.35 ± 24.25Obo7.13 ± 0.0330.00 ± 0.00189.23 ± 5.131.32 ± 0.041269.75 ± 0.787.14 ± 0.0430.00 ± 0.00202.33 ± 3.010.78 ± 0.011282.30 ± 3.25Abono7.16 ± 0.0630.00 ± 0.00217.54 ± 3.560.98 ± 0.011243.50 ± 12.737.19 ± 0.0930.00 ± 0.00199.87 ± 4.320.52 ± 0.001260.55 ± 0.35Abease7.25 ± 0.1129.75 ± 0.35170.44 ± 2.311.02 ± 0.031258.15 ± 0.927.34 ± 0.2329.50 ± 0.71188.96 ± 2.340.61 ± 0.011238.30 ± 25.17\* N = sample size, i.e. 5 samples each for groundwater and lake water for each of the 7 settlements for 3 months within each season Table 2Mean concentrations in mg/L of some trace metals in water samples collected from seven human settlements along the Bosomtwe Crater Lake in the dry and wet seasonsSeasonSettlementLake waterGroundwaterAsFePbZnCrCdNiAsFePbZnCrCdNiDry (\*N = 105)Anyinatiase\<0.010.20 ± 0.020.17 ± 0.010.12 ± 0.010.03 ± 0.00\<0.01\<0.01\<0.010.57 ± 0.410.12 ± 0.000.14 ± 0.000.01 ± 0.00\<0.01\<0.01Nkowi\<0.010.37 ± 0.080.06 ± 0.070.21 ± 0.000.01 ± 0.00\<0.01\<0.01\<0.010.28 ± 0.040.10 ± 0.010.28 ± 0.000.02 ± 0.00\<0.01\<0.01Adwafo\<0.010.22 ± 0.020.13 ± 0.010.10 ± 0.00\<0.01\<0.01\<0.01\<0.010.58 ± 0.080.17 ± 0.010.25 ± 0.000.03 ± 0.01\<0.01\<0.01Abrodwum\<0.010.28 ± 0.040.14 ± 0.030.14 ± 0.000.05 ± 0.00\<0.01\<0.01\<0.010.20 ± 0.010.16 ± 0.020.13 ± 0.000.02 ± 0.00\<0.01\<0.01Obo\<0.010.38 ± 0.050.07 ± 0.000.21 ± 0.01\<0.01\<0.01\<0.01\<0.010.42 ± 0.040.10 ± 0.010.20 ± 0.000.02 ± 0.00\<0.01\<0.01Abono\<0.010.35 ± 0.040.09 ± 0.010.13 ± 0.010.02 ± 0.00\<0.010.02 ± 0.00\<0.010.29 ± 0.050.09 ± 0.010.19 ± 0.010.03 ± 0.00\<0.010.02 ± 0.00Abease\<0.010.27 ± 0.020.16 ± 0.020.12 ± 0.000.02 ± 0.00\<0.010.02 ± 0.00\<0.010.21 ± 0.010.16 ± 0.030.12 ± 0.010.02 ± 0.00\<0.010.02 ± 0.00Wet (\*N = 105)Anyinatiase\<0.010.27 ± 0.120.11 ± 0.030.18 ± 0.030.02 ± 0.01\<0.01\<0.01\<0.011.02 ± 1.290.12 ± 0.000.15 ± 0.000.02 ± 0.00\<0.01\<0.01Nkowi\<0.010.29 ± 0.060.10 ± 0.000.12 ± 0.050.01 ± 0.00\<0.01\<0.01\<0.010.39 ± 0.070.09 ± 0.010.28 ± 0.000.01 ± 0.01\<0.01\<0.01Adwafo\<0.010.28 ± 0.030.09 ± 0.020.11 ± 0.000.02 ± 0.00\<0.01\<0.01\<0.010.53 ± 0.050.07 ± 0.090.25 ± 0.010.03 ± 0.02\<0.01\<0.01Abrodwum\<0.010.37 ± 0.050.05 ± 0.050.14 ± 0.000.02 ± 0.00\<0.01\<0.01\<0.010.32 ± 0.060.07 ± 0.080.14 ± 0.010.02 ± 0.00\<0.01\<0.01Obo\<0.011.99 ± 0.770.06 ± 0.000.16 ± 0.030.01\<0.01\<0.01\<0.010.46 ± 0.050.07 ± 0.020.15 ± 0.030.01 ± 0.01\<0.01\<0.01Abono\<0.010.49 ± 0.080.04 ± 0.020.12 ± 0.010.02 ± 0.00\<0.010.02\<0.010.37 ± 0.050.05 ± 0.020.22 ± 0.010.02 ± 0.00\<0.010.19 ± 0.05Abease\<0.010.30 ± 0.020.07 ± 0.080.09 ± 0.010.02 ± 0.00\<0.010.02\<0.010.26 ± 0.030.06 ± 0.070.16 ± 0.020.01 ± 0.00\<0.010.19 ± 0.03\* N = sample size, i.e. 5 samples each for groundwater and lake water for each of the 7 settlements for 3 months within each season
Trace metals {#Sec9}
------------
The mean levels of trace metals showed generally no significant differences between lake and groundwater or during the dry and wet seasons (Table [2](#Tab2){ref-type="table"}). The levels of the metals arsenic and cadmium, were generally small and generally below the detection limit of the instrument, Varian AAS 220 used which is 0.01 mg/L. This shows that their levels may be below those of WHO standard values of 0.01 mg/L for As, 0.003 mg/L for Cd and 0.02 mg/L for Ni.
Fe recorded mean levels in the range of 0.20 ± 0.02--0.38 ± 0.05 mg/L and 0.20 ± 0.01--0.58 ± 0.08 mg/L in lake water and groundwater respectively in the dry season for all the sampling stations. In the wet season the mean values recorded for Fe were in the range of 0.27 ± 0.12--1.99 ± 0.77 mg/L and 0.26 ± 0.03--1.02 ± 1.02 mg/L for lake water and groundwater respectively. These give rise to overall mean values for both lake and groundwater around Lake Bosomtwe exceeding the WHO standard value of 0.3 mg/L ([@CR45]). Large amount of ingested iron may results in high iron levels in the blood which lead to reaction between iron and peroxides to produce free radicals. These free radicals are highly reactive and can damage DNA, protein lipids and other cellular components. Iron typically damage cells in the heart, liver and elsewhere which can cause significant adverse effect including coma, metabolic acidosis, shock, liver failure, coagulopathy, adult respiratory distress syndrome, long-term organ damage and even death (Dina et al. [@CR9]; Alada, [@CR2]; Ferner [@CR12]).
Pb recorded mean levels in the range of 0.06 ± 0.07--0.17 ± 0.01 mg/L and 0.09 ± 0.01--0.17 ± 0.00 mg/L in lake water and groundwater respectively in the dry season for all the sampling stations. In the wet season the mean values recorded for Pb were in the range of 0.04 ± 0.02--0.11 ± 0.03 mg/L and 0.06 ± 0.07--0.12 ± 0.00 mg/L for lake water and groundwater respectively. All the recorded mean levels for Pb were above the WHO recommended standard of 0.01 mg/L. Mean Pb levels were generally relatively higher in the wet season than the dry season both lake and groundwater (Table [2](#Tab2){ref-type="table"}). Lead damages the nervous connections especially in young children and cause blood and brain disorders.
Lead poisoning typically results from ingestion of food and water contaminated with lead. Several studies have shown that chronic lead exposure reduces nerve conduction velocity in peripheral nerves in adult subjects without clinical symptom or sign of disease (Skerfring [@CR15]). Long-term exposure to lead cause small increase in blood pressure especially middle-aged and older people and can cause anaemia. Exposure to high level can severely damage the brain and kidney in males (Golub [@CR15]).
Zn recorded mean levels in the range of 0.10 ± 0.00 to 0.21 ± 0.00 mg/L and 0.12 ± 0.01--0.28 ± 0.00 mg/L in lake water and groundwater respectively in the dry season for all the sampling stations. In the wet season the mean values recorded for Zn were in the range of 0.09 ± 0.01--0.18 ± 0.03 mg/L and 0.14 ± 0.01--0.28 ± 0.00 mg/L for lake water and groundwater respectively. In general the levels of Zn in the water samples from groundwater were higher than those in the water samples from the lake. Zn levels at Nkowi and Obo were similarly higher in the wet season as compared to the dry season. Zinc is an essential mineral of exceptional biological and public health importance (Hambridge and Krebs [@CR12]). Zinc is believed to possess antioxidant properties, which may protect human beings against accelerated aging of the skin and muscles in the body. It also helps to speed up the healing process after injury (Milbury and Richer [@CR12]). Zinc toxicity can occur in both acute and chronic forms. Acute adverse effects of high intake of zinc include nausea, vomiting, loss of appetite, abdominal cramps diarrhea and headache (Prasad [@CR19]). Excessive amount of zinc is harmful as it suppresses copper and iron absorption (Fosmire [@CR9]). Zinc deficiency is associated with chronic liver disease, chronic renal disease, sickle cell disease, diabetes, malignancy and other chronic illness (Prasad [@CR19]).
Cr recorded mean levels in the range of below detection to 0.05 ± 0.00 mg/L and 0.02 ± 0.00 to 0.03 ± 0.00 mg/L in lake water and groundwater respectively in the dry season for all the sampling stations. In the wet season the mean values recorded for Cr were in the range of below detection to 0.02 ± 0.00 mg/L and 0.01 ± 0.00 to 0.03 ± 0.02 mg/L for lake water and groundwater respectively. Chromium has no verified biological role and has been classified as not essential for mammals (Bona et al. [@CR2]). Cr(VI) is very toxic and mutagenic but has not been established as a carcinogen when in solution although it may cause allergic contact dermatitis (ACD) (Bona et al. [@CR2]). Several in vitro studies indicated that high concentration of Cr(III) in the cell can lead to DNA damage (Eastmond et al. [@CR2]).
The level of Ni were generally below the detection limit of the instrument nickel except at two sampling stations where they recorded levels of 0.19 ± 0.05 mg/L and 0.19 ± 0.03 mg/L. Nickel plays an important role in the biology of the micro-organisms and plants (Astrid et al. [@CR4]). However, high concentrations of nickel have been reported to be toxic and have carcinogenic effects for a wide range of life forms (ATSDR [@CR2]; Goyer [@CR11]; IARC [@CR19]). Sensitized individuals may show allergy to nickel resulting in skins-dermatitis (skin itch). Once a person is sensitized to nickel, any further contact will produce a reaction and adverse health effects can occur at far lower concentrations compared to non-sensitized individuals (ATSDR [@CR2]).
Multivariate analysis {#Sec10}
---------------------
Figures [2](#Fig2){ref-type="fig"} and [3](#Fig3){ref-type="fig"} show the Box and Whisker plots of metals in the surface and groundwater from the Bosomtwi Crater Lake and its surroundings in the dry and wet seasons. The asterisks and white circles represent outliers. The effects of outliers on the outcome of the multivariate analysis were not significant at 95 % confidence limit by Student's t test. Tables [3](#Tab3){ref-type="table"} and [4](#Tab4){ref-type="table"} show the Principal component/factor analysis (PCA/FA) for selected trace metals in lake water and groundwater respectively for wet and dry seasons. PCA/FA identified two principal components (PC) for lake water in both wet and dry seasons with % total variance of 80.99 and 72.73 respectively. PCA/FA for groundwater identified two principal components (PC) for the wet season with % total variance of 73.80 and three principal components for the dry season with % total variance of 87.33. The ratios calculated for these trace metals suggest mineral weathering as a vital geochemical process that control the concentrations of the trace metals in both the lake water and groundwater in the study area.Fig. 2Box and whisker plots of metals in the water samples from the Bosomtwe Crater Lake in the **a** dry and **b** wet seasons Fig. 3Box and whisker plots of metals in the groundwater sample from the Bosomtwe Crater Lake in the **a** dry and **b** wet seasons Table 3Factor loading for selected trace metals in lake water samples for wet and dry seasonsVariableDry seasonWet seasonPC 1PC 2PC 1PC 2Fe−0.5410.2360.3380.677Pb0.564−0.0740.372−0.611Zn−0.531−0.2620.5390.059Cr0.318−0.139−0.406−0.292Ni0.0800.922−0.5400.282Eigenvalues2.9181.1312.2341.403% of variance58.36022.62744.67728.052Cumulative %58.36080.98744.67772.728 Table 4Factor loading for selected trace metals in groundwater samples for wet and dry seasonsVariableDry seasonWet seasonPC 1PC 2PC 3PC 1PC 2Fe0.492−0.3660.1420.577−0.005Pb−0.144−0.0690.9130.625−0.083Zn0.6350.285−0.167−0.1250.643Cr0.3660.6720.3290.0000.713Ni−0.4470.574−0.104−0.511−0.265Eigenvalues1.9351.3121.1202.3001.390% of variance38.70126.23222.39545.99827.801Cumulative %38.70164.93387.32845.99873.799
The R-mode HCA was used to determine the relationship among the various trace metals, pH, conductivity and temperature using Ward's method (Squared Euclidean distance as measure of similarity). Cluster analysis (CA) grouped these parameters into clusters on the basis of similarities within a group and dissimilarities between different groups. Parameters belonging to the same cluster are likely to have originated from a common source. The R-mode CA produced two clusters based on spatial similarities and dissimilarities (Figs. [4](#Fig4){ref-type="fig"}, [5](#Fig5){ref-type="fig"}) for the lake and groundwater for both dry and wet seasons. Cluster 1 for dry seasons for the groundwater and lake water contained the same parameters (i.e. Fe, Pb, Zn, Cr, Ni, TDS, Turbidity and Temp) whereas cluster 2 contained only pH and conductivity. Similarly for the wet seasons Cluster 1 for both groundwater and lake water contained the same parameters (Fe, Pb, Zn, Cr, Ni, TDS, Turbidity and Temp) whereas cluster 2 contained pH and conductivity. The fact that cluster 1 for both lake water and groundwater for both seasons contained Pb, Zn, Cr, Ni and Fe seem to suggest that these metals are mainly of lithogenic rather than anthropogenic sources.Fig. 4Dendrogram of trace metals pH, conductivity and temperature in the lake water samples from Bosomtwe Crater Lake in the **a** dry and **b** wet seasons using ward's method Fig. 5Dendrogram of trace metals pH, conductivity and temperature in groundwater samples from the surroundings of Bosomtwe Crater Lake in the **a** dry and **b** wet seasons using ward's method
Health risk assessment for some selected metals {#Sec11}
-----------------------------------------------
The results of non-carcinogenic health risk assessment for the selected metals in both lake and groundwater for adults and children via ingestion and dermal routes are given in Tables [5](#Tab5){ref-type="table"} and [6](#Tab6){ref-type="table"}. Human beings are exposed to trace metals through direct ingestion, inhalation through mouth and nose, and dermal absorption through skin. The health risk associated with ingestion of water depends on the volume of water consumed and the weight of the individual. Hence health risk assessment was determined using the maximum and minimum concentration of the Fe, Pb, Zn, Cr and Ni in the lake and groundwater. The results indicate that the hazard quotient through injection of water HQ~ing~ from the Bosomtwe Crater Lake for all the metals during both wet and dry seasons were less than unity for both adults and children which indicates that these metals could pose minimum hazard to local residents. The HQ~derm~ values were also found to be less than unity which means that dermal adsorption of the metals may have little or no health threat. Health hazard indices (HI) on exposure to water from both lake water and ground water through ingestion and dermal contacts for both seasons are less than unity. It is therefore obvious from the results that in both cases the observed values are below the safe limit of unity which clearly indicates that there was no cumulative potential of adverse risk in water sampled via direct ingestion or dermal ingestion to the Inhabitants. The chronic risk assessment (CDI) for the metals in the groundwater samples from habitats surrounding the Bosomtwe Crater Lake and water samples from the lake through the ingestion pathway is given in Table [7](#Tab7){ref-type="table"}. Generally health risk assessment indices like HQ and CDI and the overall non-carcinogenic health risk assessment (HI) less than unity are indicative of less significant risk through the ingestion route or dermal contact (Moore and Ramamoorthy [@CR39]; Wu et al. [@CR11]).Table 5Health risk assessment for the metals in the lake water samples from the Bosomtwe Crater Lake through ingestion and dermal absorption pathways during the wet and dry seasons for adults and childrenRfD~ing~ (µg/kg/day)RfD~derm~ (µg/kg/day)Wet seasonDry seasonAdultsChildrenAdultsChildrenHQ~ing~HQ~derm~HQ~ing~HQ~derm~HQ~ing~HQ~derm~HQ~ing~HQ~derm~Fe7001401.16 × 10^−5^--8.57 × 10^−5^2.76 × 10^−7^--2.03 × 10^−6^4.44 × 10^−5^--3.27 × 10^−4^8.14 × 10^−7^--6 × 10^−6^9.47 × 10^−5^--1.64 × 10^−5^2.25 × 10^−7^--3.88 × 10^−7^5.75 × 10^−3^--1.40 × 10^−2^6.63 × 10^−7^--1.15 × 10^−6^Pb1.40.428.61 × 10^−4^--2.37 × 10^−3^5.45 × 10^−5^--1.5 × 10^−4^3.29 × 10^−3^--9.04 × 10^−3^1.61 × 10^−4^--4.42 × 10^−4^1.51 × 10^−3^--3.66 × 10^−3^9.53 × 10^−5^--2.32 × 10^−4^3.62 × 10^−5^--6.25 × 10^−5^2.81 × 10^−4^--6.83 × 10^−4^Zn3001209.04 × 10^−6^--1.81 × 10^−5^6.44 × 10^−7^--1.29 × 10^−6^3.45 × 10^−5^--6.9 × 10^−5^1.9 × 10^−6^--3.8 × 10^−6^1 × 10^−5^--2.11 × 10^−5^7.15 × 10^−7^--1.5 × 10^−6^3.84 × 10^−5^--8.05 × 10^−5^2.11 × 10^−6^--4.43 × 10^−6^Cr30.0750--2.01 × 10^−4^0--7.63 × 10^−5^0--7.67 × 10^−4^0--2.25 × 10^−4^0--5.02 × 10^−4^0--1.91 × 10^−4^0--1.92 × 10^−3^0--5.63 × 10^−4^Ni205.40--3.01 × 10^−5^0--2.12 × 10^−6^0--1.15 × 10^−4^0--6.25 × 10^−6^0--3.01 × 10^−5^0--2.12 × 10^−6^0--1.15 × 10^−4^0--6.25 × 10^−6^HI~ing/derm~----8.82 × 10^−4^--2.70 × 10^−3^5.54 × 10^−5^--2.32 × 10^−4^3.37 × 10^−3^--1.03 × 10^−2^1.64 × 10^−4^--6.83 × 10^−4^1.53 × 10^−3^--4.23 × 10^−3^9.63 × 10^−5^--4.27 × 10^−3^5.83 × 10^−3^--1.62 × 10^−2^2.84 × 10^−4^--1.95 × 10^−3^ Table 6Health risk assessment for the metals in the groundwater samples from groundwater in settlements around the Bosomtwe Crater Lake through ingestion and dermal absorption pathways during the wet and dry seasons for adults and childrenRfD~ing~ (µg/kg/day)RfD~derm~ (µg/kg/day)Wet seasonDry seasonAdultsChildrenAdultsChildrenHQ~ing~HQ~derm~HQ~ing~HQ~derm~HQ~ing~HQ~derm~HQ~ing~HQ~derm~Fe7001401.2 × 10^−5^--4.39 × 10^−5^2.66 × 10^−7^--1.04 × 10^−6^4.27 × 10^−5^--1.68 × 10^−4^7.84 × 10^−7^--3.07 × 10^−6^8.61 × 10^−6^--2.5 × 10^−5^2.04 × 10^−7^--5.92 × 10^−7^3.29 × 10^−5^--9.53 × 10^−5^6.03 × 10^−7^--1.75 × 10^−6^Pb1.40.421.08 × 10^−4^--2.58 × 10^−3^6.81 × 10^−5^--1.63 × 10^−4^4.11 × 10^−3^--9.86 × 10^3^2.01 × 10^−4^--4.82 × 10^−4^2.94 × 10^−3^--3.66 × 10^−3^1.23 × 10^−4^--2.32 × 10^−4^7.40 × 10^−3^--1.40 × 10^−2^3.62 × 10^−4^--6.83 × 10^−4^Zn3001201.41 × 10^−5^--2.81 × 10^−5^1 × 10^−6^--2 × 10^−6^5.37 × 10^−5^--1.07 × 10^−4^2.95 × 10^−6^--5.91 × 10^−6^1.31 × 10^−5^--2.81 × 10^−5^9.3 × 10^−7^--2 × 10^−6^4.99 × 10^−5^--1.07 × 10^−4^2.74 × 10^−6^--5.91 × 10^−6^Cr30.0751 × 10^−4^--3.01 × 10^−4^3.81 × 10^−5^--1.14 × 10^−4^3.84 × 10^−4^--1.15 × 10^−3^1.13 × 10^−4^--3.38 × 10^−4^2.01 × 10^−4^--5.02 × 10^−4^7.63 × 10^−5^--1.14 × 10^−4^7.67 × 10^−4^--1.15 × 10^−3^2.25 × 10^−4^--3.38 × 10^−4^Ni205.40--2.86 × 10^−4^0--2.01 × 10^−5^0--1.09 × 10^−3^0--5.94 × 10^−5^0--3.01 × 10^−5^0--2.12 × 10^−6^1.09 × 10^−4^--1.15 × 10^−3^0--6.25 × 10^−6^HI~ing/derm~----2.33 × 10^−4^--3.24 × 10^−3^1.075 × 10^−4^--3.00 × 10^−4^4.59 × 10^−3^--1.24 × 10^−2^3.18 × 10^−4^--8.88 × 10^−4^2.16 × 10^−3^--4.24 × 10^−3^2.00 × 10^−4^--3.51 × 10^−4^9.34 × 10^−3^--1.54 × 10^−2^5.90 × 10^−4^--1.04 × 10^−3^ Table 7Chronic risk assessment (CDI) for the metals in the groundwater samples from settlements around the Bosomtwe Crater Lake and water samples from the lake through the ingestion pathway during the wet and dry seasons for adults and childrenDry seasonWet seasonLake waterUndergroundLake waterGroundwaterAdultsChildrenAdultChildrenAdultChildrenAdultChildrenFe6.29 × 10^−3^--1.19 × 10^−2^2.40 × 10^−2^--4.56 × 10^−2^6.29 × 10^−3^--1.82 × 10^−2^2.40 × 10^−2^--6.96 × 10^−2^8.49 × 10^−3^--6.24 × 10^−2^3.24 × 10^−2^--2.39 × 10^−1^8.17 × 10^−3^--3.21 × 10^−2^3.12 × 10^−2^--1.22 × 10^−1^Pb1.89 × 10^−3^--5.34 × 10^−3^7.20 × 10^−3^--2.04 × 10^−2^2.83 × 10^−3^--5.34 × 10^−3^1.08 × 10^−2^--2.04 × 10^−2^1.26 × 10^−3^--3.46 × 10^−3^4.80 × 10^−3^--1.32 × 10^−2^1.57 × 10^−3^--3.77 × 10^−3^6.00 × 10^−3^--1.44 × 10^−2^Zn3.14 × 10^−3^--6.60 × 10^−3^1.20 × 10^−2^--2.52 × 10^−2^3.77 × 10^−3^--1.07 × 10^−4^1.44 × 10^−2^--5.91 × 10^−6^2.83 × 10^−3^--5.03 × 10^−5^1.08 × 10^−2^--1.92 × 10^−2^4.40 × 10^−3^--8.8 × 10^−3^1.68 × 10^−2^--3.36 × 10^−2^Cr3.14 × 10^−4^--1.57 × 10^−3^1.20 × 10^−3^--6.00 × 10^−3^3.14 × 10^−4^--8.80 × 10^−^31.20 × 10^−3^--3.36 × 10^−3^3.14 × 10^−4^--6.29 × 10^−4^1.20 × 10^−3^--2.40 × 10^−3^3.14 × 10^−4^--9.43 × 10^−4^1.20 × 10^−3^--3.60 × 10^−3^Ni0--6.29 × 10^−4^0--2.40 × 10^−3^0--6.29 × 10^−4^0--2.40 × 10^−3^0--00--00--5.97E−030--2.28E−02
The carcinogenic risk (CR) assessment for the metals are through ingestion of both lake water and groundwater in the lake and its environs is given in Table [8](#Tab8){ref-type="table"}. The CR value has been calculated for calculated for Cr and Pb only because the value of cancer slop for Zn, Fe and Ni could not be assessed in the Integrated Risk Information System (IRIS, provided by USEPA database; USEPA [@CR19]). Generally a CR value greater than 1 in a million (10^−6^) is considered significant by USEPA. The results show that Pb and Cr exhibited ranges of carcinogenic indices exceeding 10^−6^ for both lake and groundwater for both adults and children for both seasons. This indicate that ingestion of both lake and groundwater from the lake and its environs poses carcinogenic risk with regard to the level of Pb and Cr. Hence appropriate control measures and interventions should be put in place to protect the health of the human population in the study area.Table 8Carcinogenic risk assessment (CR~ing~) for the metals in the groundwater from settlements around the Bosomtwe Crater Lake and water samples from the lake through the ingestion pathway during the wet and dry seasons for adults and childrenDry seasonWet seasonLake waterUndergroundLake waterGroundwaterAdultsChildrenAdultChildrenAdultChildrenAdultchildrenPb2.13 × 10^−4^--6.03 × 10^−4^8.12 × 10^−4^--2.30 × 10^−3^3.19 × 10^−4^--6.03 × 10^−4^1.22 × 10^−3^--2.30 × 10^−3^1.42 × 10^−4^--3.90 × 10^−4^5.41 × 10^−4^--1.49 × 10^−3^1.77 × 10^−4^--4.25 × 10^−4^6.77 × 10^−4^--1.62 × 10^−3^Cr6.03 × 10^−7^--3.01 × 10^−6^2.3 × 10^−6^--1.15 × 10^−5^6.03 × 10^−7^--1.81 × 10^−6^2.3 × 10^−6^--6.9 × 10^−6^6.03 × 10^−7^--1.21 × 10^−6^2.3 × 10^−6^--4.6 × 10^−6^6.03 × 10^−7^--1.81 × 10^−6^2.3 × 10^−6^--6.9 × 10^−6^
Conclusion {#Sec12}
==========
The mean levels of Fe and Pb were above the WHO values for both lake and groundwater for both wet and dry seasons whereas the mean levels of Zn and Cr were below the WHO values for both lake and groundwater for both seasons. The hazard quotients (HQ) and health hazard indices (HI) through ingestion and dermal contact of lake and groundwater in towns around the lake for both adults and children gave values which were below the acceptable limit (\< 1), indicating the absence of non-carcinogenic health risk to the communities. The study however reveals that ingestion of both lake and groundwater from the lake and its surroundings poses carcinogenic risk with regard to the level of Pb and Cr. Hence appropriate control measures and interventions should be put in place to protect the health of the human population in the study area.
CDI
: Chronic risk assessment
CR
: Carcinogenic risk
WHO
: World Health Organization
HQ
: Hazard quotient
AAS
: Atomic absorption spectrometer
HI
: Hazard index
IRIS
: Integrated Risk Information System
TAB performed the laboratory work. NKA-D and AAA conceptualized this study. TAB and NKA-D drafted the manuscript. NKA-D, and AAA provided the guidance and reviewed the manuscript. All authors read and approved the final manuscript.
Acknowledgements {#FPar1}
================
The authors are grateful to the National Council for Tertiary Education (NTCE), Ghana for a research grant under the Teaching and Learning Innovation Fund (TALIF-KNUSTS/3/005/2005). We are also grateful to AngloGold Ashanti (Ghana) limited for the use of their Atomic Absorption Spectrophotometer.
Competing interests {#FPar2}
===================
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction
===============
RM 8640 consists of six plastic bottles each containing a 2 mL suspension of polymethyl methacrylate (PMMA) microspheres with a specified amount of immobilized fluorescein isothiocyanate (FITC). The microspheres are intended for calibrating the fluorescence response of flow cytometers \[[@b1-j110-2gai]\]. This paper describes the procedures used for assigning values of *MESF* (molecules of equivalent soluble fluorophore) to the microspheres with immobilized FITC. There are three major measurements in this procedure. First, the concentration of microspheres is measured using a Multisizer 3 (Coulter Corporation, Miami FL) particle counter[1](#fn1-j110-2gai){ref-type="fn"}. Second, a fluorometer is calibrated using SRM 1932, a fluorescein solution. Third, the fluorescence signal is measured for each of the microsphere suspensions. Finally, the data from the three measurements are used to calculate the *MESF* values of the microspheres.
In practice, we made serial dilutions of SRM 1932 and calibrated the fluorometer response as a function of fluorescein concentration. Since the concentration of fluorescein varies from 10^−12^ mol/L to 10^−9^ mol/L, it was necessary to pay special attention to contamination, linearity, photodegradation, and background subtraction. The measurements of the concentration of the microsphere suspension constitutes the operational definition of particle concentration. There are at present no particle number standards to validate the concentration measurement.
A cytometer was used to measure the fluorescence signals associated with the five microsphere populations each with a different amount of immobilized FITC. A valid assignment of *MESF* values should yield a linear relation between the measured fluorescence signal in the cytometer and the assigned *MESF* values. A linearization procedure was used to impose a linear relation between the cytometer response of the five microsphere populations and their *MESF* values.
2. Revised Measurement Model and the Assignment of *MESF*
=========================================================
It was pointed out \[[@b2-j110-2gai]\] that the quantum yield as defined in Eqs. (A3) and (A14) in Ref. \[[@b3-j110-2gai]\] is the ratio of fluorescent radiant flux to absorbed radiant flux. Since the radiant flux is the product of the number flux and average spectral energy, the ratio of radiant fluxes is not the same as the ratio of number fluxes. In the following we introduce a modification to the measurement model which allows a consistent use of quantum yield, a molecular property defined in terms of number flux.
In the previous paper \[[@b3-j110-2gai]\] we expressed the fluorescence spectral radiance, *L*~f~(*λ*~m~,*λ*~x~), as $$\begin{array}{l}
{L_{f}\left( {\lambda_{m},\lambda_{x}} \right) = \frac{S_{f}\left( \lambda_{m} \right)}{S^{\prime}\left( \lambda_{m} \right)}L^{\prime}\left( \lambda_{m} \right)} \\
{= 2.3FNly\left( {\lambda_{m},\lambda_{x}} \right)\varepsilon\left( \lambda_{x} \right)\Phi_{i}\left( \lambda_{x} \right)\Delta\lambda_{x}} \\
\end{array}$$
*λ*~m~ and *λ*~x~ are the emission and excitation wavelengths, *S*~f~, *S*′, are measured signals, *L*′ is the spectral radiance of a reference source, *F*, *l*, *Φ*~i~(*λ*~x~), and ∆*λ*~x~ are instrument characteristics, and *N*, *y*, and *ε* are sample properties. The quantity *y*(*λ*~m~, *λ*~x~) relates the absorbed radiant flux at wavelength *λ*~x~ to the fluorescent radiant flux at wavelength *λ*~m~. In other words, the radiant flux from fluorescence is a fraction *y* of the absorbed radiant flux. The radiant flux can be converted to a photon number flux by dividing the radiant flux by the energy of a single photon. Thus the quantity *y*(*λ*~m~, *λ*~x~) can be converted into a relation between fluorescence photon number flux and absorbed photon number flux by multiplying it by the ratio of the respective wavelengths. $$y^{\prime}\left( {\lambda_{m},\lambda_{x}} \right) = \frac{\lambda_{m}}{\lambda_{x}}y\left( {\lambda_{m},\lambda_{x}} \right).$$
The quantity *y*′(*λ*~m~, *λ*~x~) is conveniently separated into a quantum yield *φ* and a normalized relative photon emission function *s*′(*λ*~m~, *λ*~x~) (1/nm), where $${\int{s^{\prime}\left( {\lambda_{m},\lambda_{x}} \right)d\lambda_{m}}} = 1.$$
Thus [Eq. (1)](#fd1-j110-2gai){ref-type="disp-formula"} can be rewritten as $$\frac{S_{f}\left( \lambda_{m} \right)}{S^{\prime}\left( \lambda_{m} \right)}\frac{\lambda_{m}}{\lambda_{x}}L^{\prime}\left( \lambda_{m} \right) = 2.3FNl\phi s^{\prime}\left( {\lambda_{m},\lambda_{x}} \right)\varepsilon\left( \lambda_{x} \right)\Phi_{i}\left( \lambda_{x} \right)\Delta\lambda_{x}.$$
Integrating over all emission wavelengths gives $${\int{S_{f}\left( \lambda_{m} \right)\left\lbrack {\frac{\lambda_{m}}{\lambda_{x}}\frac{L^{\prime}\left( \lambda_{m} \right)}{S^{\prime}\left( \lambda_{m} \right)}} \right\rbrack}}d\lambda_{m} = \Omega N\phi\varepsilon\left( \lambda_{x} \right)I_{0}.$$
The change in the measurement model amounts to a modification of the spectral correction function given by the term in brackets on the left side of [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"}. [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"} supplants a similar equation in Ref. \[[@b3-j110-2gai]\]. The left side of [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"} is found experimentally by performing the specified operation on the measured fluorescence emission spectrum. The reference source used to calibrate the detector wavelength response was unpolarized. In order to minimize the effects due to polarization of the fluorescence emission, the incident laser beam polarization was confined to the plane defined by the incident and detected light beams. We made no estimate of possible artifacts due to polarization differences.
Suppose that the operation in [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"} is performed on the emission spectrum from a reference solution with known concentration of fluorophore and the emission spectrum from a suspension of microspheres with immobilized FITC. Furthermore, the number concentration of microspheres is known. In the case that the two numbers on the left of [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"} are equal, the corresponding solution and suspension properties on the right side of [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"} are also equal. $$\mathrm{\Omega}N_{\text{sol}}\phi_{\text{sol}}\varepsilon_{\text{sol}}\left( \lambda_{x} \right)I_{0} = \mathrm{\Omega}N_{\text{sus}}\phi_{\text{sus}}\varepsilon_{\text{sus}}\left( \lambda_{x} \right)I_{0}.$$
The subscripts "sol" and "sus" in [Eq. (6)](#fd6-j110-2gai){ref-type="disp-formula"} refer to solution and suspension, respectively. We assume that the experimental conditions used for the measurements on solution and suspension are the same and that the solution and suspension have equal indexes of refraction. In that case, *I*~0~ and Ω are the same on both sides of [Eq. (6)](#fd6-j110-2gai){ref-type="disp-formula"} and can be factored out. $$N_{\text{sol}}\phi_{\text{sol}}\varepsilon_{\text{sol}}\left( \lambda_{x} \right) = N_{\text{sus}}\phi_{\text{sus}}\varepsilon_{\text{sus}}\left( \lambda_{x} \right).$$
It is known from measurements that there are shifts in the wavelength of maximum absorption. It is likely that the absolute value of the extinction coefficient is also different. The extinction coefficient of fluorophore immobilized on a microsphere has not been measured due to predominance of scattering. However, we make the major assumption that the molar extinction coefficient is the same for fluorophore in solution and immobilized on the microsphere. Thus, [Eq. (7)](#fd7-j110-2gai){ref-type="disp-formula"} reduces to an equality of fluorescence yields. $$N_{\text{sol}}\varphi_{\text{sol}} = N_{\text{sus}}\phi_{\text{sus}}.$$
Based on [Eq. (8)](#fd8-j110-2gai){ref-type="disp-formula"}*N*~sol~ is equivalent to *N*~sus~. The calculation used to assign molecules of equivalent soluble fluorophore (*MESF*) values to the microspheres is given by $$MESF = \frac{N_{A}}{1000}\frac{C_{\text{eq}}}{N_{\text{spheres}}}$$where *N*~sus~ = *N*~sphere~ (mL^−1^) is the number concentration of fluorescein labeled microspheres and *C*~eq~ is the molar concentration of soluble fluorescein which gives the same value for the left side of [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"} as the suspension of microspheres. Avogadro's constant, *N*~A~ (6.022 × 10^23^), in [Eq. (9)](#fd9-j110-2gai){ref-type="disp-formula"} is a conversion factor between molar and number concentrations. The equivalent concentration of soluble fluorescein is determined using the fluorescein calibration curve $$C_{\text{eq}} = 10^{- \text{intercept}}\left( {FS \times P_{\text{adj}}} \right)^{\text{slope}}$$where "intercept" and "slope" are the linear fit parameters describing the relationship between the logarithm of the observed fluorescence signal and the logarithm of the concentration of fluorescein. *FS* is the fluorescence signal of the microsphere suspension evaluated according to the left side of [Eq. (5)](#fd5-j110-2gai){ref-type="disp-formula"}. The value *FS* has to be adjusted for possible differences in illumination intensity between the calibration measurements and the microsphere measurements. The ratio of the average of the power readings taken during the calibration and microsphere measurements is set equal to the adjustment factor, *P*~adj~, which multiplies *FS* in [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"}. The power readings varied by less than 1 % during the calibration or the microsphere measurements. However, since the two measurements were taken on different days, the difference in average power could be as high as several percent. [Eq. (9)](#fd9-j110-2gai){ref-type="disp-formula"} and [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"} constitute the operational definition of the *MESF* assignment.
In the following we describe the procedure used for obtaining the five factors needed in [Eq. (9)](#fd9-j110-2gai){ref-type="disp-formula"} and [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"}. As an example, the value of *MESF* is obtained for one specific suspension of microspheres.
3. Calibration of the Fluorometer
=================================
[Figure 1](#f1-j110-2gai){ref-type="fig"} shows a schematic of the fluorometer used in the *MESF* assignments. A water-cooled argon ion laser (Lexel model 95) was the source of 488 nm light. A glass slide reflected a portion of the output beam and directed it towards a photo diode (Newport 818 UV) whose output was processed by a power meter (Newport 1815-C). The accuracy of the power reading was about 2 % (product specifications). The output of the power meter was monitored as an index of relative illumination power which was used to determine the factor *P*~adj~ in [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"}. The laser beam, transmitted through the glass slide, passed a neutral density (ND) filter wheel which contained ND filters of nominal optical density (OD) values 0, 1, 2, and 3. The beam passed through a laser line filter to eliminate plasma lines from the laser and then was directed via two mirrors to the entrance aperture of a 10× microscope objective. The two reflections changed the vertical polarization into horizontal. The objective lens focused the laser beam on a square capillary flow cell (WWP100375 from Polymicro Technologies) with a inner dimension of 100 µm. The portion of the capillary that was illuminated by the laser was stripped of the polyimide coating which normally surrounds the glass capillary. The capillary flow cell was mounted on a rotary stage which in turn was mounted on a *X*-*Y* translation stage. The rotary stage was used to position the capillary so that the incident beam was perpendicular to one of the sides. The *X*-*Y* stage was used to position the capillary at the position of the monochromator entrance slit image.
The Model 270M monochromator, made by JY Horiba, was equipped with a CCD-3000 Detector System. The CCD chip used in the measurements was back illuminated with a minimum quantum efficiency of 65 % at 550 nm. The operational temperature of the chip was 213 K with a dark current less than 4 e^−^ pixel^−1^ min^−1^. The pixel layout was 1024 by 256, and the dynamic range of the 16 bit analog to digital converter was 65535 digital number (*DN*). The measurements were performed by binning the 256 pixels in the short direction of the CCD chip. The entrance slit of the monochromator was equipped with a mechanical shutter, the single axial Model 227MCD (JY Horiba). Appendix A describes the procedures used to validate the performance of the fluorometer \[[@b4-j110-2gai]\].
3.1 Calibration of the Fluorometer With SRM 1932
------------------------------------------------
SRM 1932 certifies the concentration of fluorescein as (60.97 ± 0.88) µmol/kg. Given the density of the SRM buffer as 1.003 g/mL, the molar concentration of the SRM 1932 is (61.15 ± 0.88) µmol/L. This value of the concentration was used as the initial concentration. Serial dilutions were made by combining previously made solution with additional buffer. All solutions were prepared gravimetrically using a calibrated balance (Sartorius 2024MP) with a resolution of 0.01 mg. The errors were obtained from the standard deviation of four weighins. The standard error in the fluorescein concentrations was about 1.6 % and originated mostly from the initial error in the SRM concentration. The weighin errors contributed a minimal error to the final fluorescein concentrations.
3.2 Measurement of the Fluorescence Signal
------------------------------------------
The fluorescein solution was pumped with a peristaltic pump through a capillary flow cell \[[@b5-j110-2gai]\]. The 10× objective and the capillary were mounted on appropriate mounts to provide the necessary adjustment. A good adjustment was indicated by the appearance of a clean circular beam cross section after the transit through the capillary. Poor adjustment was characterized by complex interference bands from light reflected by various surfaces of the capillary. The capillary tube (length = 0.7 m) was coupled to a plastic tube (length = 20 cm) via a stainless steel pressure "Swagelok" coupling. The plastic tube was inserted into the pumping mechanism of a peristaltic pump (PP). A centrifuge vial (1.5 mL capacity) contained the test solution and a small magnetic stirring bar. The ends of the capillary or the plastic tube were inserted in the solution and the pump direction set appropriately to pump the solution to the waste container. Neither the capillary nor the plastic tube touched the waste solution. Formation and detachment of small droplets above the waste container was an indicator of flow.
The spectrum was accumulated over a period of 40 s leading to appearance of sharp spikes in the CCD response. The spikes are pixels with unusually large charge content, the spikes are confined to one or two adjacent pixels. We removed the spikes by simply replacing the contents of the pixel containing a spike (attributed to cosmic ray events hitting a pixel in the CCD array) by an average of the contents of two nearby pixels. The necessary overall dynamic range was obtained by varying the integration time from 0.5 s to 40 s, and by the dynamic range of the CCD itself (about 60). The ND filter was not used to modify the illumination intensity.
The peristaltic pump produced a pulsating flow with a time period of approximately 1.6 s. The strong illumination caused substantial photodegradation of the fluorescein solution in the capillary. The photodegradation and the pulsating flow produced a time variation in the fluorescence signal. This variation was averaged adequately during integration times longer then 5 s. However, for shorter integration times we had to take multiple measurements and average the resulting fluorescence signals. The pulsating fluorescence intensity enlarged the measurement errors for the concentrated fluorescein solutions. We avoided changing the intensity of the illuminating beam since that would change the photodegradation rate for measurements performed on solutions with different fluorescein concentrations. The assumption is made that the photodegradation is the same for solution and fluorescein immobilized on microspheres. To minimize possible systematic errors due to photodegradation, the flow conditions for the calibration and microsphere measurements were made as similar as possible. The polarization anisotropy for fluorescein solutions was approximately zero. Therefore, the measured spectra are characteristic of a solution of random emitters and systematic effects due to polarization are small.
3.3 Background Subtraction
--------------------------
[Figure 2a](#f2-j110-2gai){ref-type="fig"} shows the measured spectra of a pure phosphate buffer (solid circles) and a fluorescein solution with a concentration of approximately 16 pM (open circles) in phosphate buffer. The laser line filter shown in [Fig. 1](#f1-j110-2gai){ref-type="fig"} was a critical component in the fluorometer since without it the spectra in [Fig. 2a](#f2-j110-2gai){ref-type="fig"} would be an order of magnitude higher and dominated by laser plasma lines. The spectra in [Fig. 2a](#f2-j110-2gai){ref-type="fig"} have *DN* values larger than 2000, suggesting that CCD linearity correction was not significant. [Figure 2b](#f2-j110-2gai){ref-type="fig"} shows the spectrum when the buffer response is subtracted from the fluorescein solution response. As expected, the emission spectrum peaks at approximately 510 nm. The integration of the spectrum was performed by summing the *DN* values of the subtracted spectrum in [Fig. 2b](#f2-j110-2gai){ref-type="fig"}. The resulting truncation errors for wavelengths less than 595 nm and wavelengths greater than 620 nm were not evaluated. We estimate that these truncations lead to a systematic bias of less than −1 %. Similar spectra were collected for solutions with higher values of fluorescein concentration and the integration time was lowered as required to insure that the resulting CCD response was not saturated. The subtracted spectrum was corrected for the spectral response of the detector as described in Appendix A. The corrected integration times were used to normalize all integrated fluorescence signals (*FS*) to the fluorescence signal (*FS*) at an integration time of 1 s.
[Figure 3](#f3-j110-2gai){ref-type="fig"} shows a log-log plot of the integrated *DN* values on the horizontal axis and five different concentrations of fluorescein plotted on the vertical axis. The best linear fit to the data in [Fig. 3](#f3-j110-2gai){ref-type="fig"} is $$\log(\text{concentration}) = - 15.20 + 0.957 \times \log(FS)$$
The errors of the fit parameters were 0.09 and 0.009 for the intercept and slope, respectively. The errors were obtained from a linear regression procedure in Mathcad. The lowest point (for a concentration of approximately 7 pmol/L) was systematically lower than expected from the linear trend set by the higher points. We rationalize this as an effect of adsorption of fluorescein on the capillary walls \[[@b4-j110-2gai]\]. A calibration was accepted if the slope of the best fit fell between 0.95 and 1.05. A perfectly linear relation has a slope of 1.0, however a deviation of ± 0.05 was accepted. The values of "intercept" and "slope" are used in [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"}.
4. Measurement of Fluorescence From Microspheres With Immobilized FITC
======================================================================
The measurements of fluorescence signal were carried out in the identical apparatus as the calibration with serially diluted fluorescein solutions. Since the microsphere measurements were carried out after the fluorescein solution, great care was needed to eliminate possible contamination. Prior to the microsphere measurements, the capillary flow cell was washed for several hours. The washing was performed by pumping buffer through the capillary in alternating directions. The switch in pumping direction was important to clear possible dead spaces in the connection between the capillary tube and the plastic tube. As a rule of thumb, the cleaning was sufficiently good when the CCD signal was about 370 *DN* at 510 nm with an integration time of 10 s, and a power indicator of about 20. This number was obtained through experience. During all measurements the following pumping sequence was followed: first the suspension was pumped through the plastic tube into the capillary (this direction filled the capillary quickly); second the suspension was pumped through the capillary into the tube and fluorescence spectra were accumulated. Normal flow was indicated by the presence of a scattering diffraction pattern in the transmitted light. The pattern indicated the presence of spheres in the capillary as well as proper alignment. The intensity of the diffraction pattern fluctuated as expected since the number of microspheres in the sensing volume was of the order of twenty. On occasion, small bubbles passed through the illuminated region in the capillary. The passage of a bubble was obvious from the distortion of the transmitted laser beam. Clogging was obvious because flow stopped and the fluorescence signal decreased.
4.1 Fluorescence Measurement
----------------------------
[Figure 4a](#f4-j110-2gai){ref-type="fig"} shows the spectrum measured for a suspension of microspheres (Suspension \#1, open circles) and a suspension of blank microspheres (solid circles). The suspensions were identical to those used in the microsphere concentration measurements. The measurements in [Fig. 4a](#f4-j110-2gai){ref-type="fig"} were performed with integrating time of 40 s, and power indicator displaying 19.2. [Figure 4b](#f4-j110-2gai){ref-type="fig"} shows the difference spectrum. The location of maximum emission shifts to the red, and the spectrum is broader. Both facts are typical of emission from immobilized FITC and serve as additional indicators that the flow cell is clean and free of fluorescein in solution. The quality of background subtraction was gauged by the disappearance of the water Raman line centered at 585 nm. In further analysis, the spectra in [Fig. 4a](#f4-j110-2gai){ref-type="fig"} were corrected for CCD non linearity (≈0.8 % effect) and normalized to an integration time of 1 s. The spectrum was summed and the errors of the summed spectrum were estimated from multiple measurements. Spectral response corrections were made on the data in [Fig. 4](#f4-j110-2gai){ref-type="fig"} (≈3 % effect). The resultant summed spectrum for Suspension \#1 was (11220 ± 600) *DN*. The value of the summed spectrum was substituted for FS in [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"}. Taking the ratio of average power levels during calibration and microsphere fluorescence measurements gave *P*~adj~ = 0.92 ± 0.02.
5. Measurement of Microsphere Concentration
===========================================
The microsphere concentration was measured using a Coulter Multisizer 3 particle counter. The instrument detects small changes in conductivity between two reservoirs separated by a narrow orifice of diameter 100 µm. Whenever a particle passes through the orifice from one reservoir to the other, a transient change in conductivity is detected and recorded as a particle. In practice, the reservoir outside the orifice is the test reservoir into which samples are placed. A volumetric syringe withdraws fluid from the test reservoir through the orifice and the concentration is determined as the particle counts divided by the preset volume of fluid withdrawn. The fluid that was placed into the test reservoir was obtained from the original fluid by diluting the original fluid about 500 times using Isotone fluid. Isotone is a proprietary fluid which has optimal properties for the performance of the Multisizer 3. The composition of Isotone is sufficiently close to that of phosphate buffer saline (PBS) so that microsphere suspension stability is not degraded. The concentration of the particles in the original fluid is found by multiplying the measured concentration by the reciprocal of the dilution.
5.1 Determination of the Dilution Factor of the Test Suspension
---------------------------------------------------------------
The microspheres were obtained from Bangs Laboratories as a special order and were delivered in 5 mL opaque plastic bottles. The nominal solids mass fraction was about 2 %. The microspheres, as shipped, were suspended in a solvent optimized for enhancing the stability of the suspension. For the purpose of measuring the fluorescence signal of the microspheres, it was necessary to re-suspend the microspheres in pH = 7.2 PBS. The first step was to take about 1 ml of the microsphere suspension directly from the plastic bottle and centrifuge it (Marathon 13K centrifuge from Fisher Scientific at 2000 rpm). The pellet of microspheres was then re-suspended in 1 ml of PBS, pH 7.2, containing ≈0.1 mg SDS. An additional advantage of buffer exchange is that possible soluble fluorescent impurities in the bottle are minimized. The mass of about 200 µg of the resulting suspension was placed into a container with a mass of Isotone equivalent to about 100 ml of Isotone. The mass of the buffer was determined using a balance, OHaus ED4130, while the suspension was transferred using calibrated 100 µL or 200 µL pipettes and the mass approximated as 0.1996 g using the known density of water. A gravimetric check of the amount of solution delivered by the 100 µL pipette gave (0.10016 ± 0.0006) g using water at 23 °C and a Sartorius 2024 MP balance. The error in the reciprocal of the dilution factor was estimated to be less than 1 %. The remaining suspension was used for fluorescence measurements as described above.
5.2 Concentration Measurement
-----------------------------
Prior to each series of measurements the operation of the Multisizer 3 was verified by running a suspension of calibration microspheres. The quality of the measurement was evaluated by measuring the mean diameter of the microspheres, and the coefficient of variation (CV) of the diameter values. The diluted microsphere suspension was placed in the Multisizer 3 and stirred at an indicated rate of 13 rpm. An image of the orifice was always examined to verify that the orifice was free from debris. The instrument analytical volume was set to 500 µL or 1 ml and a dilution factor (as determined above) was entered as a parameter in the acquisition software. [Figure 5](#f5-j110-2gai){ref-type="fig"} shows a typical result of a single Multisizer 3 measurement. The horizontal axis gives the inferred particle diameter, and the vertical axis gives the number of times a particle with this diameter passed through the orifice. Using the dilution factor, the vertical axis can be converted into concentration. The inferred concentration of Suspension 1 was (1.427 ± 0.016) × 10^6^ mL^−1^ for all particles with diameters between 6.3 µm and 8.1 µm. The peak to the right of the main peak corresponds to "doublets". These are events associated with the passage of two particles. The concentration of "doublets" (diameters between 8.3 µm and 9.7 µm) was calculated to be (0.077 ± 0.002) × 10^6^ mL^−1^. Since the concentration of particles in the test solution is small, it is likely that the "doublets" correspond to permanent associations of two microspheres. For permanent associations, the radii of doublets and triplets are related to the particle radius by *r*~doublet~ = 2^1/3^*r*~sphere~ and *r*~triplets~ = 3^1/3^*r*~sphere~ respectively. These relations are consistent with the observed size distributions of singlets, doublets, and triplets. This measurement of particle associations is relevant to the proper interpretation of the fluorescence signal as discussed below. The concentration measurement was repeated five times and the error associated with each concentration was found by dividing the standard deviation by the square root of 5, the number of trials. It should be noted that the error in the concentration is about 5 % implying that the error due to the uncertainty of the dilution factor (about 1 %) is negligible. The error of the concentration measurement is due mainly to the statistics of the particles entering the orifice. Whenever possible, the errors in the concentration measurement were minimized by using more concentrated test suspensions and larger sampling volumes.
5.3 Correction for Microsphere Doublets
---------------------------------------
Both the Multisizer 3 and the cytometer measurements show the presence of microsphere doublets- permanent associations of two microspheres. Some aggregation of colloidal particles is expected and the amount depends on previous treatments (e.g., sonication, age, and solvent). The concentration measurements were performed with the same suspension as the fluorescence measurements. Furthermore the two measurements were performed on the same day. The simplest assumptions are that the Multisizer 3 gives the correct concentration of doublets in the suspension used for fluorescence measurements and that the fluorescence intensity from doublets is twice the fluorescence intensity from single microspheres. Therefore, in comparing the fluorescence yield of a suspension and solution we multiply the doublet concentration by a factor of two relative to the singlet concentration. $$N_{\text{Total}} = N_{\text{Single}} + 2N_{\text{Double}}.$$
Where *N*~Single~ is the number concentration of single microspheres, *N*~Double~ is the number concentration of double microspheres, and *N*~Total~ is the total concentration which should be used in comparing fluorescence yields in [Eq. (9)](#fd9-j110-2gai){ref-type="disp-formula"}. The assumption that the fluorescence intensity from a double microsphere is twice the fluorescence from a single microsphere is qualified by such consideration as distortion of the illuminating light and changes in quantum yield at point of contact between the microspheres. The measured polarization anisotropy in microsphere suspensions was about 0.08. This small value was rationalized by the large tether length of the chain of seven carbon atoms that immobilized the FITC to the microsphere surface. A long tether permits considerable rotational freedom. The systematic effects due to polarization differences between solution and microsphere suspension were neglected.
6. Assignment of *MESF* Values
==============================
The assignment of values of molecules of equivalent soluble fluorophore (*MESF*) entails the comparison of the fluorescence signal from suspensions of microspheres with immobilized fluorophore and solutions of fluorophore \[[@b3-j110-2gai]\]. The comparison requires the integration of the emitted fluorescence over all wavelengths. The comparison requires a correction for differences in the molar absorption coefficient of soluble and immobilized fluorescein. Fluorescence excitation spectra show that the absorption spectra of the soluble and immobilized fluorescein are shifted relative to each other \[[@b3-j110-2gai]\]. The fluorescence excitation spectra allow us to estimate the change in absorbance at 488 nm due to the shift. However, we were not able to measure the absolute value of the molar absorption coefficient for the immobilized fluorescein. This adjustment is left for a future refinement. The values of FS, "intercept", "slope", Padj, and *N*~spheres~ were used in [Eq. (9)](#fd9-j110-2gai){ref-type="disp-formula"} and [Eq. (10)](#fd10-j110-2gai){ref-type="disp-formula"} to calculate a value of *MESF* of 1667 ± 400. The same procedure was carried out for the other suspensions. Four independent series of measurements were carried out and the resulting *MESF* values averaged to give the final value reported in the Certificate of Investigation.
7. Linearization of *MESF* Assignments With a Cytometer
=======================================================
The objective of this measurement was to demonstrate that the *MESF* assignments obtained with the fluorometer were self-consistent. The microspheres were passed through a cytometer, and the response was measured. Each population of microspheres produced a population of fluorescence pulses which are characterized by a mean pulse area and a standard variation. Since the cytometer response is linear, the mean pulse height of the five populations of microspheres should correlate linearly with the assigned values of *MESF*. The cytometer does not provide a measurement of the absolute *MESF* values, however it does place a stringent constraint on relative *MESF* values. The *MESF* values obtained with the fluorometer were modified so that they correlate linearly with the mean channel measured with the cytometer. Appendix B gives details of the procedure used for validating the performance of the cytometer.
7.1 Microsphere Measurements
----------------------------
Alignment microspheres from Spherotech Corp. were used to align the cytometer laser beam. The alignment was sufficiently good when the fluorescence pulse mean channel was between 180 000 *DN* and 200 000 *DN*, and the CV was better than 4 %. After the alignment, the six populations of the microspheres in the reference material were mixed and pumped through the cytometer and the corresponding fluorescence and scattering peaks recorded. The suspensions were prepared by putting two drops of the suspension from each of the six plastic bottles into 1 ml of PBS buffer. [Figure 6](#f6-j110-2gai){ref-type="fig"} shows typical results. Note that the single and double microsphere signals are resolved. The dense groupings with circular bounds correspond to single microsphere signals. Diagonally to the upper right of each dense grouping are less dense groupings (not enclosed by boundaries) corresponding to the passage of double microspheres. The table in [Fig. 6](#f6-j110-2gai){ref-type="fig"} gives the properties of the groups of dots enclosed by the circular bounds. The *Y* geometric mean gives the mean scattering pulse amplitude, while the *X* geometric mean gives the mean fluorescence pulse height for each population. As expected, the scattering is relatively constant for the five populations, while the fluorescence signal differs substantially.
[Figure 7](#f7-j110-2gai){ref-type="fig"} shows a plot of the log of the mean fluorescence channel (*X* geometric means in [Fig. 6](#f6-j110-2gai){ref-type="fig"}) associated with each microsphere as a function of the log of the assigned *MESF* value for each microsphere. As described above, the *MESF* assignments were performed using the fluorometer. A fit to a straight line gives a slope of 0.984, indicating that there is a small deviations from linearity. All points are within two standard deviations of the best straight line fit. Since the cytometer is a linear device, the measured mean channels should correlate linearly with the *MESF* values which are proportionate to the number of fluorophore on the microsphere and hence to the fluorescence signal. This fact provides a means to linearize the assigned *MESF* values.
7.2 Linearization
-----------------
The response of the cytometer was shown to be linear over the dynamic range encompassing the response of the five microsphere populations. Therefore, the *MESF* values have to correlate linearly with the observed mean channels in the cytometer measurement. The cytometer measurement provides a relative ordering of the *MESF* values but gives no measure of the absolute values. The procedure that was used to impose a linear relation on the *MESF* values determined by fluorometer was as follows. The solid circles in [Fig. 7](#f7-j110-2gai){ref-type="fig"} show the *MESF* values obtained in Series 4 measurements as a function of mean cytometer channel. The data in [Fig. 7](#f7-j110-2gai){ref-type="fig"} was fitted with a straight line whose slope was constrained to 1.027, which characterizes the cytometer linearity. Next, the *MESF* value given by the straight line was calculated for each mean channel. This calculated value is the linearized *MESF* value obtained for that series of measurements.
8. Certification of *MESF* Values
=================================
The *MESF* values were assigned by averaging the values obtained in four independent measurements each consisting of 1) calibration of the fluorometer using SRM 1932, 2) measurement of the microsphere fluorescence intensity, and 3) measurement of the microsphere concentration. The data in each set of four measurements gave an assignment of *MESF* values to the five populations of microspheres. The four different assignments provide a measure of reproducibility and an estimate of random error. The possible error obtained from the variation of the four *MESF* assignments was consistent with the error estimate for each of the four *MESF* assignments. The average values are reproduced in the Certificate of Investigation for RM 8640. The cytometer measurements were used to linearize the *MESF* values yielding values of linearized *MESF*. The four values of linearized *MESF* were averaged to give the average linearized *MESF* values in the Certificate of Investigation. The certificate includes both the average *MESF* values determined by the fluorometer, and the average linearized *MESF* values.
The *MESF* values were assigned under certain assumptions which are restated below in the order of decreasing relevance. The molar extinction coefficient is the same for fluorescein in solution and fluorescein immobilized on the microspheres. It is known that there are differences in the wavelength of maximum absorption between fluorescein in solution and on the surface of the microsphere. It is expected that the magnitude at maximum absorption will also be different, however these have not been measured yet. The difference in molar extinction coefficient can be taken into account as soon as the values become available. The equality of fluorescence yield would imply that *N*~sus~ microspheres are equivalent to a concentration of soluble fluorophore given by *N*~sol~\[*ε*~sol~(*λ*~x~)/*ε*~sus~(*λ*~x~)\].Photodegradation of fluorescein is the same in solution and at the surface of the microsphere. Systematic measurements of photodegradation in the two environments are not available. Differences in photodegradation rate would lead to systematic differences in the fluorescence signal between calibration measurements and microsphere measurements.Residual polarization of fluorescence emission is the same for fluorescein in solution and at the surface of the microsphere. The measured polarization anisotropies are slightly different for fluorescein in solution and on the surface of a microsphere. Sensitivity of detectors to polarization would lead to small systematic differences in fluorescence signal between fluorescein in solution and on the surface of microspheres.Adsorption on capillary flow cell walls is the same. If the adsorption (and holdup in dead spaces) of fluorescein in solution and fluorescein on microspheres is not the same, then there will be systematic differences in the fluorescence signal at low and high concentrations.Illumination of FITC immobilized on microspheres is the same as the illumination of fluorescein in homogeneous solution.
9. Conclusions
==============
A method was described for comparing the fluorescence yields of a solution of fluorescein and a suspension of microspheres with immobilized fluorescein isothiocyanate (FITC). The equality of fluorescence yields leads to an assignment of molecules of equivalent soluble fluorophore (*MESF*) to a microsphere with immobilized FITC. The *MESF* values may be the appropriate units for comparing fluorescence measurements. There is a need to investigate the validity of some of the assumptions that were made in carrying out the *MESF* assignment as described above.
10. Appendix A. Fluorometer Characterization
============================================
10.1 Wavelength Calibration
---------------------------
The wavelength calibration was performed using mercury lamp lines, and a neon lamp (Oriel 6032). The parameters in the data acquisition software were adjusted so that over the wavelength range 480 nm to 700 nm, the deviation between the true and measured wavelengths was less than 1 nm.
10.2 CCD Linearity
------------------
The output of a light emitting diode (LED) was passed through a variable ND filter and focused on the end of a bifurcated optical fiber. One of the arms of the optical fiber was placed at the location of the sample in the monochromator setup (see [Fig. 1](#f1-j110-2gai){ref-type="fig"}), and the other arm was placed in front of a photo diode (PD). Silicon photodiodes are known to be linear within 0.1 % up to a photocurrent of 200 µA \[[@b6-j110-2gai]\]. The LED spectrum was recorded by the CCD for an integration time of 0.2 s. The spectrum was taken with automatic subtraction of the dark current. The CCD response was characterized by finding the average value of *DN* in a narrow range (518 nm to 522 nm) of wavelengths around the maximum response. The integrated CCD response was compared to the photo diode reading.
We examined the dependence of the average signal in *DN* as a function of the PD response. There was a substantial deviation from linearity at low values of *DN*. The most likely source of this deviation is trapping of electrons during the readout process \[[@b7-j110-2gai]\]. During readout, the electrons are switched sequentially from pixel to pixel. The trapping centers prevent some of the electrons from reaching the final register pixel. Since the number of trapping sites is fixed, the relative importance of these sites increases with decreasing number of total electrons. The measured deviations provide a factor which can be used to linearize the CCD response. Thus the linearized *DN* value is given by $$\begin{array}{l}
{{(DN)}_{\text{linear}} = {(DN)}_{\text{measured}}(1 + f)} \\
{\mspace{67mu} f = 10^{\lbrack 0.6418 - 0.7181^{\prime}\log(DN_{\text{measured}})\rbrack}} \\
\end{array}$$where (*DN*)~measured~ is the *DN* value that is read out during a measurement and (*DN*)~linear~ is the linearized value of the CCD response which is used in further data analysis. The factor *f* was obtained from the fit of the deviation of the measured CCD data from the linear PD response. We did not assign an error to the correction. The correction was applied to background measurements and sample measurements prior to subtraction of background. The correction is negligible for *DN* values above 600.
10.3 Integration Time Linearity
-------------------------------
The CCD accumulates electrons for a preset integration time that is determined by a mechanical shutter located after the entrance slit of the monochromator. To measure the correspondence between the integration time setting in the software and the actual time, we illuminated the monochromator entrance slit with a constant light source and measured the CCD response for different indicated integration times *t*~indicated~. The reference light source (see Sec. 10.4) was used as the constant light source. The CCD response was integrated from 540 nm to 560 nm. The ratio of the CCD response divided by indicated time *t*~indicated~ relative to the CCD response for 1 s of indicated integration time showed a bias in the indicated time setting. At indicated times longer than 1 s, the actual integration time is shorter than the indicted time. The difference between the actual integration time, *t*~actual~, and the indicated integration time, *t*~indicated~, was corrected by multiplying the indicated time by the correction factor 1.00102--0.000782 × *t*~indicated~.
10.4 Spectral Response
----------------------
The reference lamp's output was calibrated at NIST over the range of wavelengths 340 nm to 800 nm in steps of 20 nm. The output port of the calibrated lamp was placed at the location of the sample in the fluorometer. The variable iris of the calibrated lamp was set so that the CCD response was between 10 000 *DN* and 40 000 *DN*. The CCD linearity is excellent in this region.
We formed the ratio of calibrated output of the reference lamp to the output at 520 nm (normalized reference output) and compared it to the ratio of measured CCD response of the reference lamp to the CCD response at 520 nm (normalized CCD response). The sharp decline in the response at shorter wavelengths was due to the holographic filter which was used to reject the 488 nm excitation light. The spectral correction factor was obtained by dividing the normalized reference output by the normalized CCD response to the reference lamp. Multiplying the measured CCD response by the correction factor corrects for the variability of the detector response over the wavelength range.
11. Appendix B. Cytometer Characterization
==========================================
The cytometer was constructed to be as simple as possible with all physical processes open to inspection. An air-cooled Argon ion laser (Omnichrome Model 150) provided the source of 488 nm illumination. The laser beam was focused by a spherical lens with a focal length of 50 mm. The focal point was located in the flow channel of a cytometer flow cell provided by Becton Dickinson Biosciences. The sample was pumped by a syringe pump (Yale Apparatus Model YA-12), and the sheath fluid was pumped by the pressure in the container vessel. A flow meter (Aalberg Model TMR1-010426) in the sheath flow line gave an indication of the flow rate (usually set to 90 scale units). The laser beam passed through a glass plate whose orientation provided a sensitive adjustment of the beam position in the scattering plane. The flow cell contained the collection optics that focused the emitted light about 25 cm from the lens where photomultiplier (PMT) detectors (Hamamatsu Model H6780) were placed. The usual arrangement of dichroic mirror (DM) and bandpass filters selected the fluorescence and elastically scattered light components. The outputs from the two PMTs were processed by digital electronics provided by Becton Dickinson Biosciences (BD FACSDiVa system). An oscilloscope provided a visual monitor of the pulses associated with the side scattering (SSC) and the first fluorescence (FL1) channels of the detection electronics.
11.1 Linearity and Dynamic Range
--------------------------------
The output of a green LED was focused on a slit of a chopper and then split by a glass plate reflector and passed to the inputs of two optical fiber (FO) bundles. One of the split beams (that transmitted through the glass plate) passed through a neutral density (ND) filter, and the FO routed the light to the PMT associated with the first fluorescence channel (FL1). The other beam (reflected from the glass plate) was incident on the FO that routed the light to the PMT associated with the side scattering channel (SSC). The chopper rotation was adjusted to give pulses that approximated the duration of the pulses from the microspheres in the flow cell. The adjustment was performed by visual inspection of the oscilloscope traces. The pulse rate was about 190 pulses per second. The linearity measurements were performed by noting the mean channel of the detected pulses in FL1 for a given OD value of the ND filter. Six ND filters were purchased from Newport Corp., and the attenuation values were used as provided by the manufacturer. The data was accumulated by recording events for different OD values of the ND filters. The six mean channels corresponding to OD values of 0, 0.51, 1.05, 1.50, 1.98. and 2.49 were recorded. The mean channels were plotted on a log-log scale versus the OD values. The average slope was found to be 1.027 ± 0.008. Therefore, we conclude that the cytometer response is linear. We assume that the photon pulses in the above simulation and from fluorescent microspheres behave in an identical fashion.
The noise properties of the cytometer are given by the coefficient of variation (CV) defined as the standard deviation of a pulse distribution divided by the mean pulse amplitude \[[@b8-j110-2gai]\]. A linear relation between (CV)^2^ and the inverse of the mean pulse amplitude was observed, and indicated that the main source of noise was the statistics of photon arrival at the photomultiplier cathode.
The authors are indebted to Dr. John Lu for assistance in the statistical analysis of data. We are indebted to Dr. Gary Kramer for the use of a reference light source.
Certain commercial equipment, instruments, or materials are identified in this paper to foster understanding. Such identification does not imply recommendation or endorsement by NIST, nor does it imply that the materials or equipment identified are necessarily the best available for the purpose.
**About the authors:** A. K. Gaigalas is a physicist and Lili Wang a research chemist both in the Biotechnology Division of the NIST Chemical Science and Technology Laboratory. Abe Schwartz is a chemist at the Center for Quantitative Cytometry. Gerald E. Marti is a review and research officer at CBER, FDA and an attending physician at NCI. Robert F. Vogt is a chemist at the Center of Disease Control and Prevention, Atlanta, GA. The National Institute of Standards and Technology is an agency of the Technology Administration, U.S. Department of Commerce.
{#f1-j110-2gai}
{#f2-j110-2gai}
{#f3-j110-2gai}
{#f4-j110-2gai}
{#f5-j110-2gai}
{#f6-j110-2gai}
{ref-type="fig"}). The cytometer measurements provide a relative relation between the *MESF* values. The *MESF* values determined by the fluorometer were modified slightly to conform to the linear relation as given by the cytometer measurements.](j110-2gaif7){#f7-j110-2gai}
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
hnRNP A1 is an RNA-binding protein that contains two RNA-binding domains (RBDs) and a glycine-rich domain responsible for protein--protein interaction. It is involved in pre-mRNA splicing and transport of cellular RNAs (reviewed by [@cdd462c9]). It is predominantly located in the nucleus, but also shuttles between the nucleus and the cytoplasm ([@cdd462c35]). The signal that mediates shuttling has been identified as a 38 amino acid sequence, termed M9, located near the C-terminus of hnRNP A1 between amino acids 268 and 305 ([@cdd462c32]; [@cdd462c40]; [@cdd462c48]). Yeast two-hybrid screening with M9 as bait resulted in the discovery of a novel transportin-mediated pathway for nuclear import of hnRNP A1 ([@cdd462c36]; [@cdd462c11]; [@cdd462c41]). The function of the cytoplasmic hnRNP A1 has not been well defined. Studies have shown that cytoplasmic and nuclear hnRNP A1 exhibit different RNA-binding profiles. Cytoplasmic hnRNP A1 is capable of high-affinity binding to AU-rich elements that modulate mRNA turnover and translation ([@cdd462c14], [@cdd462c15]; [@cdd462c17]). It has also been shown to promote ribosome binding to mRNAs by a cap-mediated mechanism, and prevent spurious initiation at aberrant translation start sites ([@cdd462c43]).
MHV belongs to the Coronaviridae family of positive-sense, single-stranded RNA viruses. MHV replication and transcription occur exclusively in the cytoplasm of infected cells via the viral RNA-dependent RNA polymerase (RdRp) (reviewed by [@cdd462c24]). Initially, the 5′-most gene 1 of the viral genome is translated into the viral RdRp, which then replicates the viral genomic RNAs into negative-strand RNAs. Subsequently, the negative-strand RNAs are used as templates to transcribe mRNAs, which include a genomic-sized RNA and a nested set of subgenomic mRNA transcripts, all with an identical 5′ non-translated leader sequence of 72--77 nucleotides and 3′ co-terminal polyadenylated ends. The subgenomic mRNA transcription of MHV utilizes a unique discontinuous mechanism in which the leader sequence, often derived from a different molecule, is fused to RNAs at the intergenic (IG) sites (i.e. transcription initiation site) to generate subgenomic mRNAs ([@cdd462c21]; [@cdd462c26]; [@cdd462c51]). The exact mechanism of how these mRNAs are made is still controversial. However, it has been shown that the process of discontinuous RNA transcription is regulated by several viral RNA elements, including the *cis*- and *trans*-acting leader RNA ([@cdd462c26]; [@cdd462c51]), IG sequence ([@cdd462c30]) and 3′-end untranslated sequence ([@cdd462c27]). There is considerable biochemical evidence suggesting possible direct or indirect interactions between the various RNA regulatory elements. hnRNP A1 binds MHV negative (--)-strand leader and IG sequences ([@cdd462c12]; [@cdd462c25]). Site-directed mutagenesis of the IG sequences demonstrated that the extent of binding of hnRNP A1 to the IG sequences correlated with the efficiency of transcription from the IG site ([@cdd462c50]; [@cdd462c25]). Immunostaining of hnRNP A1 showed that hnRNP A1 relocated to the cytoplasm of MHV-infected cells, where viral RNA synthesis occurs ([@cdd462c25]). hnRNP A1 also mediates the formation of a ribonucleoprotein complex containing the MHV (--)-strand leader and IG sequences ([@cdd462c52]). These results suggest that hnRNP A1 may serve as a protein mediator for distant RNA regions to interact with each other.
Many cellular proteins, including calreticulin ([@cdd462c39]), polypyrimidine tract-binding protein (PTB) ([@cdd462c16]; [@cdd462c49]), La protein ([@cdd462c33]), Sam68 ([@cdd462c31]), poly(rC)-binding protein ([@cdd462c34]) and nucleolin ([@cdd462c45]), have been implicated to be involved in viral RNA transcription or replication. In addition to MHV, hnRNP A1 has also been reported to interact with human cytomegalovirus immediate-early gene 2 protein, which plays an important role in the regulation of virus replication ([@cdd462c47]). Furthermore, a yeast protein related to human core RNA splicing factors, Lsm1p, has been shown to be required for the efficient replication of brome mosaic virus RNA ([@cdd462c8]). Recently, Reddy and colleagues demonstrated an inhibition of HIV replication by dominant-negative mutants of Sam68 ([@cdd462c37]). However, none of these cellular proteins has been shown experimentally to participate directly in RNA-dependent RNA synthesis.
In order to demonstrate the involvement of hnRNP A1 in MHV RNA replication and transcription, we established several DBT cell lines stably expressing either the wild-type (wt) hnRNP A1 or a C-terminus-truncated mutant lacking the M9 sequence and part of the glycine-rich domain. We showed that the mutant hnRNP A1, which was localized predominantly in the cytoplasm, exhibited dominant-negative effects on viral genomic RNA replication and subgenomic mRNA transcription. In contrast, overexpression of the wt hnRNP A1 accelerated the synthesis of all viral RNAs. Our results provide strong evidence that hnRNP A1 is directly or indirectly involved in MHV RNA synthesis in the cytoplasm and that the C-terminal part of the protein is important for its function. This finding thus reveals a novel function for hnRNP A1 in the cytoplasm.
Results
=======
Characterization of stable cell lines expressing the wt and a C-terminus-truncated hnRNP A1
-------------------------------------------------------------------------------------------
To explore a potential role for hnRNP A1 in MHV RNA synthesis, we established murine DBT cell lines stably expressing the Flag-tagged wt hnRNP A1 (DBT-A1) or a mutant hnRNP A1, which has a 75 amino acid deletion from the C-terminus (DBT-A1ΔC) (Figure [1](#cdd462f1){ref-type="fig"}A). This mutant lacks part of the glycine-rich domain and the M9 sequence responsible for shuttling hnRNP A1 between the nucleus and the cytoplasm. Immunoblot of the whole-cell lysates with an anti-Flag antibody detected a 34 kDa protein in DBT-A1 cells and a 27 kDa protein in three independent clones of DBT-A1ΔC cells (Figure [1](#cdd462f1){ref-type="fig"}B), whereas no protein was cross-reactive to the anti-Flag antibody in the control cell line stably transfected with the pcDNA3.1 vector (DBT-VEC). The amounts of the Flag-tagged wt and truncated hnRNP A1 were comparable in these cell lines. A chicken polyclonal antibody against hnRNP A1 detected two endogenous hnRNP A1 isoforms or hnRNP A1-related proteins in the whole-cell lysates of all of the cell lines. The bottom band (34 kDa) overlaps the Flag-tagged wt hnRNP A1 in DBT-A1 cells. There was only a slight increase in the overall amount of hnRNP A1 in DBT-A1 cells as compared with DBT-VEC cells, indicating that the exogenous hnRNP A1 constituted a small fraction of the total hnRNP A1 in the cells. In DBT-A1ΔC cells, an additional band of smaller size (27 kDa) corresponding to the mutant hnRNP A1 was detected. The overall expression levels of the exogenous hnRNP A1 and hnRNP A1ΔC were ∼3-fold lower than that of the endogenous hnRNP A1 in whole-cell lysates (Figure [1](#cdd462f1){ref-type="fig"}B). Similar to the endogenous hnRNP A1 protein ([@cdd462c35]), the Flag-tagged wt hnRNP A1 was localized almost exclusively in the nucleus (Figure [1](#cdd462f1){ref-type="fig"}C). The mutant hnRNP A1, however, was localized predominantly in the cytoplasm (Figure [1](#cdd462f1){ref-type="fig"}C), consistent with the previous finding that the M9 nuclear localization signal is necessary to localize hnRNP A1 to the nucleus ([@cdd462c40]; [@cdd462c48]). Thus, hnRNP A1ΔC was much more abundant than the endogenous hnRNP A1 in the cytoplasm. The expression levels of the wt or mutant hnRNP A1 varied among individual cells based on immunofluorescent staining (Figure [1](#cdd462f1){ref-type="fig"}C). The growth rate (Figure [1](#cdd462f1){ref-type="fig"}D) and cell morphology (data not shown) were similar among the different cell lines.
The effects of overexpression of the wt and mutant hnRNP A1 on syncytium formation and virus production
-------------------------------------------------------------------------------------------------------
We first assessed the effects of hnRNP A1 overexpression on the morphological changes induced by MHV-A59 infection using several different clones of DBT cell lines. Virus infection was performed at a multiplicity of infection (m.o.i.) of 0.5 to detect the subtle morphological differences among the different cell lines. Syncytia appeared at ∼7 h post-infection (p.i.) in DBT-VEC cells and ∼1 h earlier in DBT-A1 cells. At both 8 and 14 h p.i., syncytia were significantly larger and more spread out in DBT-A1 cells than those in DBT-VEC cells (Figure [2](#cdd462f2){ref-type="fig"}A). Similar differences were observed with two additional clones of DBT-A1 cells (data not shown). In contrast, no syncytium was observed in three different clones of DBT-A1ΔC cells, even at 14 h p.i. At 24 h p.i., almost all DBT-A1 cells detached from the plate, but ∼10--20% of DBT-VEC cells still remained on the plate (data not shown). Remarkably, there was no sign of syncytium formation in DBT-A1ΔC cells until 24 h after virus infection, when the overall morphology of the cells was similar to that of DBT-VEC cells at 7 h p.i. (data not shown). All of the DBT-A1ΔC cells were eventually killed at ∼48 h p.i., suggesting that the inhibition of viral replication was not a result of the disruption of the MHV receptor. Correspondingly, virus production from these cell lines was significantly different. Between 6 and 14 h p.i., virus production from DBT-A1ΔC cells was ∼100- to 1000-fold less than that from DBT-VEC and DBT-A1 cells (Figure [2](#cdd462f2){ref-type="fig"}B). DBT-A1 cells produced twice as many viruses as those from DBT-VEC cells during that time period.
Relocalization of hnRNP A1 during MHV infection
-----------------------------------------------
MHV RNA synthesis occurs exclusively in the cytoplasm of infected cells. In order for hnRNP A1 to participate directly in viral transcription, it has to be recruited to the site of RNA synthesis. Although hnRNP A1 shuttles between the nucleus and the cytoplasm in normal cells ([@cdd462c35]), the level of cytoplasmic hnRNP A1 is very low. We have demonstrated previously that hnRNP A1 relocates from the nucleus to the cytoplasm of MHV-infected cells ([@cdd462c25]). To determine whether the overexpressed hnRNP A1 may participate in MHV RNA synthesis, we performed immunostaining experiments using an anti-Flag antibody to localize Flag-tagged hnRNP A1. In DBT-A1 cells, a significant increase in the cytoplasmic level of hnRNP A1 and a corresponding decrease of nuclear hnRNP A1 were observed in virus-infected cell syncytia at 7 h p.i. (Figure [3](#cdd462f3){ref-type="fig"}B); these cells express the MHV nucleocapsid (N) protein in the cytoplasm (Figure [3](#cdd462f3){ref-type="fig"}A). By comparison, in the uninfected cells, which did not have N protein staining, hnRNP A1 was predominantly localized to the nucleus (arrow in Figure [3](#cdd462f3){ref-type="fig"}B). In DBT-A1ΔC cells, very few cells were stained positive for the MHV N protein at 7 h p.i. (Figure [3](#cdd462f3){ref-type="fig"}C). Significantly, the viral N protein was detected only in the cells that were stained weakly or not at all for Flag-hnRNP A1 (Figure [3](#cdd462f3){ref-type="fig"}D), suggesting that the expression of a high level of hnRNP A1ΔC interfered with viral replication.
The effects of wt and mutant hnRNP A1 on MHV protein production
---------------------------------------------------------------
We further investigated the effects of the wt and mutant hnRNP A1 on the production of MHV structural and non-structural proteins. Cytoplasmic protein was extracted from infected cell lines at different time points after infection for immunoblot analysis to detect an open reading frame (ORF) 1a product, p22 ([@cdd462c28]) and the N protein. p22 expression in DBT-VEC cells was clearly detected at 6 h p.i. and peaked at ∼16 h p.i. (Figure [4](#cdd462f4){ref-type="fig"}A). In DBT-A1 cells, p22 appeared at 5 h p.i. and peaked at ∼8 h p.i. In DBT-A1ΔC cells, no p22 protein was detected until 16 h p.i. Similar patterns of differences were observed for the N protein in these three cell lines. Actin levels in different cell lines remained relatively constant throughout the infection, except that, in DBT-A1 cells, actin was not detected at 16 and 24 h p.i. due to the loss of the dead cells (Figure [4](#cdd462f4){ref-type="fig"}A). These results clearly demonstrated that overexpression of the wt hnRNP A1 accelerated viral protein production, whereas expression of the mutant hnRNP A1 delayed it. We also performed immunofluorescent staining of the N protein at 7 h p.i. to further confirm the western blot results. As represented by images shown in Figure [4](#cdd462f4){ref-type="fig"}B, there were more DBT-A1 cells stained positive for the N protein than DBT-VEC cells. Very few cells were found to express the N protein in DBT-A1ΔC cells. The p22 and N proteins appeared as doublets in some of the lanes of Figure [4](#cdd462f4){ref-type="fig"}A, but the results varied from experiment to experiment. The N protein is known to be phosphorylated ([@cdd462c42]). Whether p22 is post-translationally modified is not known.
The effects of the wt and mutant hnRNP A1 on MHV RNA synthesis
--------------------------------------------------------------
We next examined the effects of the wt and mutant hnRNP A1 on MHV RNA synthesis. MHV-infected cell lines were labeled for 1 h with \[^3^H\]uridine in the presence of 5 µg/ml actinomycin D at different time points after infection. \[^3^H\]uridine incorporation in DBT-VEC cells increased gradually from 4 h p.i. and peaked at ∼16 h p.i. (Figure [5](#cdd462f5){ref-type="fig"}A). DBT-A1 cells showed a significantly higher level of \[^3^H\]uridine incorporation, which peaked at ∼8 h p.i. DBT-A1ΔC cells did not show any detectable level of incorporation of the radioactivity. These results suggest that hnRNP A1 regulates MHV RNA synthesis. We further assessed the production of genomic and subgenomic MHV RNAs in these cell lines by northern blot analysis. The genomic and the six subgenomic RNA species were detected at 8 h p.i. in both DBT-VEC and DBT-A1 cells; there were significantly higher steady-state levels of all of the RNA species in DBT-A1 cells (Figure [5](#cdd462f5){ref-type="fig"}B). In contrast, no viral RNA was detected in DBT-A1ΔC cells at that time point. At 16 h p.i., MHV RNA levels in DBT-VEC and DBT-A1 cells decreased generally because of the loss of the dead cells, while the smaller subgenomic RNAs became detectable in DBT-A1ΔC cells. By 24 h p.i., most viral RNA species became detectable in DBT-A1ΔC cells (Figure [5](#cdd462f5){ref-type="fig"}B, lane 10), while most of the DBT-A1 cells were dead (lane 9). These results confirmed that the synthesis of all of the viral RNA species is accelerated by overexpression of the wt hnRNP A1 and delayed by a dominant-negative mutant of hnRNP A1.
In this analysis, we also detected an additional RNA species (arrow in Figure [5](#cdd462f5){ref-type="fig"}B), which was determined to be a defective-interfering (DI) RNA by northern blot analysis using a probe representing the 5′-untranslated region (without the leader), which is present only in genomic and DI RNAs (data not shown). Interestingly, this DI RNA was inhibited to a greater extent than other RNA species in DBT-A1ΔC cells. This result suggests that the replication of DI RNAs is more sensitive to the dominant-negative inhibition by cytoplasmic hnRNP A1.
hnRNP A1ΔC inhibits transcription and replication of MHV DI RNAs
----------------------------------------------------------------
To demonstrate further that MHV RNA transcription machinery is defective in cells expressing the mutant hnRNP A1, we studied transcription of an MHV DI RNA, 25CAT, which contains a transcription promoter (derived from the IG sequence for mRNA 7, IG7) and a chloramphenicol acetyltransferase (CAT) reporter gene ([@cdd462c26]). CAT activity can be expressed from this DI RNA only if a subgenomic mRNA containing CAT sequences is produced ([@cdd462c26]). The 25CAT RNA was transfected into MHV-A59-infected cells 1 h after infection. At 8 h p.i., CAT activity in DBT-A1 cells was significantly higher than that in DBT-VEC cells (Figure [6](#cdd462f6){ref-type="fig"}A). On the other hand, CAT activity was very low in DBT-A1ΔC cells. At 24 h p.i., CAT activity in DBT-A1 cells became slightly lower than that in DBT-VEC cells because of the loss of the dead DBT-A1 cells. The CAT activity in DBT-A1ΔC was still significantly lower than that in DBT-VEC or DBT-A1 cells. These results established that mRNA transcription from the DI RNA was also inhibited by hnRNP A1ΔC.
The results shown above (Figure [5](#cdd462f5){ref-type="fig"}B) also suggest that DI RNA replication is more sensitive to the inhibitory effects of the hnRNP A1 mutant. To confirm this result, we further studied replication of another DI RNA during serial virus passages. DBT cells were infected with MHV-A59 and transfected with DIssE RNA derived from JHM virus ([@cdd462c29]); the virus released (P0) was passaged twice in DBT cells to generate P1 and P2 viruses. DBT cells were infected with these viruses, and cytoplasmic RNA was extracted for northern blot analysis using glyoxalated RNA for a better resolution of smaller RNAs. For DBT-A1ΔC cells, RNA was extracted at 36 h p.i. since viral RNA synthesis was delayed in this cell line. Cells infected with P0 viruses did not yield detectable amounts of DIssE, but contained the naturally occurring A59 DI RNA, whose replication was inhibited more strongly than the synthesis of MHV genomic and subgenomic RNAs in DBT-A1ΔC cells (Figures [5](#cdd462f5){ref-type="fig"}B, lanes 8--10 and [6](#cdd462f6){ref-type="fig"}B, lanes 1--3). However, this A59 DI RNA was not detectable in cells infected with P1 and P2 viruses (Figure [6](#cdd462f6){ref-type="fig"}B, lanes 4--9). In contrast, DIssE appeared in cells infected with P1 viruses and further increased in cells infected with P2 viruses, indicating that the replication of the smaller DIssE may have an inhibitory effect on the replication of the larger A59 DI RNA ([@cdd462c20]). Similar to the A59 DI RNA, the replication of DIssE RNA was much more strongly inhibited than that of MHV genomic and subgenomic RNAs in DBT-A1ΔC cells (Figure [6](#cdd462f6){ref-type="fig"}B, lanes 6 and 9). Our results thus suggest that MHV DI RNA replication is more dependent on the function of cytoplasmic hnRNP A1.
The mechanism of dominant-negative inhibition by the C-terminal deletion mutant of hnRNP A1
-------------------------------------------------------------------------------------------
To understand the underlying mechanism of the inhibition of MHV RNA transcription by the C-terminal-deletion mutant of hnRNP A1, we first examined the RNA- and protein-binding properties of this mutant protein. Electrophoretic mobility shift assay demonstrated that hnRNP A1ΔC retained the ability to bind the MHV (--)-strand leader RNA and to form multimers with itself, similar to the wt hnRNP A1 (data not shown); this is consistent with the fact that both of its RBDs are intact (Figure [1](#cdd462f1){ref-type="fig"}A). Furthermore, UV-crosslinking experiments showed that increasing amounts of purified glutathione *S*-transferase (GST)--hnRNP A1ΔC efficiently competed with the endogenous hnRNP A1 for the binding of the MHV (--)-strand leader RNA (Figure [7](#cdd462f7){ref-type="fig"}A), indicating that the binding of hnRNP A1ΔC to RNA was not affected. These results suggest that the RNA-binding properties of hnRNP A1ΔC were intact.
We next examined the protein-binding properties of hnRNP A1ΔC. Since hnRNP A1 has been shown to interact with the N protein, which also participates in MHV RNA synthesis ([@cdd462c7]; [@cdd462c46]), we first determined whether the dominant-negative mutant of hnRNP A1 retained the ability to interact with the N protein *in vitro*. GST pull-down assay using various truncation mutants of hnRNP A1 showed that the N protein bound the N-terminal domain (aa 1--163) of hnRNP A1 (Figure [7](#cdd462f7){ref-type="fig"}B); thus, the binding of hnRNP A1ΔC \[equivalent to hnRNP A1(1--245)\] to the N protein was not affected. We next examined the *in vivo* interaction of the wt and mutant hnRNP A1 with an MHV ORF 1a product, p22, which has been shown to co-localize with the *de novo* synthesized viral RNA (S.T.Shi and M.M.C.Lai, unpublished results) and associate with the viral replicase complex ([@cdd462c13]). Cyto plasmic extracts from MHV-A59-infected cells were immunoprecipitated with anti-Flag antibody-conjugated beads, followed by western blotting with a rabbit polyclonal antibody against p22. At 8 h p.i., p22 was co-precipitated with the Flag-tagged hnRNP A1 from DBT-A1 cells, whereas no precipitation of p22 was observed in DBT-VEC cells (Figure [7](#cdd462f7){ref-type="fig"}C). For DBT- A1ΔC cells, co-immunoprecipitation was performed at 24 h p.i., when abundant MHV proteins were synthesized. p22 was shown to co-precipitate with hnRNP A1ΔC, indicating that hnRNP A1ΔC still formed a complex with the viral polymerase gene product. These results suggest that the ability of hnRNP A1ΔC to interact with the N and polymerase proteins was not altered.
We next investigated whether the mutant hnRNP A1 is deficient in the interaction with any other cellular proteins in this RNA--protein complex. We labeled proteins in MHV-infected cells or mock-infected cells at different time points after infection and immunoprecipitated with the anti-Flag antibody. Significantly, a cellular protein of ∼250 kDa was shown to be associated only with the wt hnRNP A1, but not the mutant hnRNP A1 (Figure [7](#cdd462f7){ref-type="fig"}D), suggesting that hnRNP A1 binds to this protein through its C-terminal domain. We propose that this cellular protein is another important component of the MHV RNA transcription/replication complex.
Discussion
==========
There is an accumulating body of evidence signifying the importance of cellular factors in RNA synthesis of RNA viruses (reviewed by [@cdd462c23]). Previous studies have shown that hnRNP A1 binds to the *cis*-acting sequences of MHV template RNA and that this interaction correlates with the transcription efficiency of viral RNA *in vivo* ([@cdd462c50]; [@cdd462c25]). In addition, hnRNP A1 is also implicated in viral RNA replication by the recent finding that hnRNP A1 interacts with the 3′-ends of both positive- and negative-strand MHV RNA (P.Huang and M.M.C.Lai, unpublished results). However, the functional importance of hnRNP A1 in viral RNA synthesis has so far not been directly demonstrated. In the present study, we established that MHV RNA transcription and replication were enhanced by overexpression of the wt hnRNP A1 protein, but inhibited by expression of a dominant-negative hnRNP A1 mutant in DBT cell lines. Our results suggest that hnRNP A1 is a host protein involved in the formation of a cytoplasmic transcription/replication complex for viral RNA synthesis. This represents a novel function for hnRNP A1 in the cytoplasm.
Our results indicate that the inhibitory effects on MHV replication exhibited by the dominant-negative mutant of hnRNP A1 were relatively more prominent than the enhancement effects by overexpression of the wt hnRNP A1. This is consistent with the subcellular localization patterns of the wt and mutant hnRNP A1 proteins. The overexpressed exogenous wt hnRNP A1 in DBT-A1 cells was predominantly localized in the nucleus, similar to the endogenous hnRNP A1 (Figure [1](#cdd462f1){ref-type="fig"}C). The C-terminal-deletion mutant, however, was localized mainly in the cytoplasm. Thus, the level of hnRNP A1ΔC was much higher than the endogenous wt hnRNP A1 in the cytoplasm of DBT-A1ΔC cells, where MHV replication occurs. This result explains why hnRNP A1ΔC could have a strong dominant-negative inhibitory effect, despite the fact that it was expressed at a lower level than the endogenous hnRNP A1 (Figure [1](#cdd462f1){ref-type="fig"}B).
The effects of the expression of the wt and mutant hnRNP A1 on virus production (Figure [2](#cdd462f2){ref-type="fig"}B), viral protein synthesis (Figure [4](#cdd462f4){ref-type="fig"}A) and viral RNA synthesis (Figure [5](#cdd462f5){ref-type="fig"}A) correlated with each other. Furthermore, hnRNP A1ΔC caused not only a global inhibition of genomic RNA replication and subgenomic mRNA transcription, but also a preferential inhibition of at least two DI RNA species. These results suggest that the inhibition of MHV replication by the hnRNP A1 mutant was most likely a direct effect on viral RNA synthesis rather than an indirect effect on other aspects of cellular or viral functions. Since hnRNP A1 binds directly to the *cis*-acting MHV RNA sequences critical for MHV RNA transcription ([@cdd462c25]) and replication (P.Huang and M.M.C.Lai, unpublished results), it is most likely that hnRNP A1 may participate in the formation of the transcription/replication complex. Indeed, our data show that hnRNP A1 interacts directly or indirectly with the N protein and a gene 1 product, p22, both of which are probably associated with the viral transcription/replication complex ([@cdd462c7]; [@cdd462c46]; [@cdd462c13]). hnRNP A1 may participate directly in viral RNA synthesis in a similar role to that of transcription factors in DNA-dependent RNA synthesis, e.g. by maintaining favorable RNA conformation for RNA synthesis. Alternatively, hnRNP A1 may modulate MHV RNA transcription or replication by participating in the processing, transport and controlling the stability of viral RNAs. It has been reported that RNA processing of retroviruses, human T-cell leukemia virus type 2 ([@cdd462c2]) and HIV-1 ([@cdd462c3]), is altered by the binding of hnRNP A1 to the viral RNA regulatory elements. It is also possible that hnRNP A1 may participate in MHV RNA synthesis indirectly by affecting the production of other host cell proteins, which may, in turn, regulate MHV RNA synthesis. Since hnRNP A1 is a dose-dependent alternative splicing factor ([@cdd462c6]), even small changes in the intracellular level of hnRNP A1 can alter the splicing of other cellular proteins. Regardless of the mechanism, our study established the importance of cellular factors in viral RNA-dependent RNA synthesis.
The transcription from 25CAT RNA was strongly inhibited by the dominant-negative mutant of hnRNP A1, as shown by CAT assays (Figure [6](#cdd462f6){ref-type="fig"}A). In addition, the replication of the naturally occurring A59 DI RNA and the artificial DIssE RNA was completely abolished (Figure [6](#cdd462f6){ref-type="fig"}B). Surprisingly, the replication of MHV DI RNAs suffered a stronger inhibition by the dominant-negative mutant of hnRNP A1 than the synthesis of MHV genomic and subgenomic RNAs, suggesting that DI RNA replication may be more dependent on hnRNP A1. Although DI RNAs contain all of the *cis*-acting replication signals that are essential for their replication in normal cells ([@cdd462c22]), the small size of DI RNA may cause it to require more hnRNP A1 to maintain a critical RNA structure. It has been shown that different DI RNAs require different *cis*-acting signals for RNA replication ([@cdd462c22]).
Our results demonstrate that the C-terminal domain of hnRNP A1, including the M9 sequence and the glycine-rich region, is important for MHV RNA transcription and replication, but the mechanism of the dominant-negative effects of hnRNP A1ΔC is still not clear. hnRNP A1ΔC retains the RNA-binding and self-association ability and is capable of binding the viral proteins N and p22, which are associated with the transcription/replication complex. It is possible that hnRNP A1ΔC is not productive due to its inability to interact with other viral or cellular proteins that are involved in MHV RNA synthesis. We have found a protein of ∼250 kDa that binds only the wt, but not the mutant hnRNP A1 (Figure [7](#cdd462f7){ref-type="fig"}D). It remains to be shown whether this cellular protein is involved in MHV RNA synthesis.
In our preliminary study, we found that MHV could replicate in an erythroleukemia cell line, CB3, which was reported to lack detectable hnRNP A1 expression as a result of a retrovirus integration in one allele and loss of the other allele ([@cdd462c1]). Since hnRNP A1 protein is involved in a variety of important cellular functions, including RNA splicing, transport, turnover and translation, it is conceivable that other redundant gene products may substitute for the function of hnRNP A1 in CB3 cells. Indeed, UV-crosslinking assays using CB3 cell extracts detected two proteins comparable to hnRNP A1 in size that could interact with the MHV negative-strand leader RNA (data not shown). These proteins may represent hnRNP A1-related proteins, since many of such hnRNPs exist in the cells ([@cdd462c5]; [@cdd462c4]). Therefore, multiple cellular proteins may have the capacity to be involved in MHV RNA synthesis.
Based on previous findings ([@cdd462c22]; [@cdd462c50]; [@cdd462c25]) and the results from this study, we propose a model for the regulation of transcription/replication of MHV RNA by hnRNP A1. We hypothesize that hnRNP A1 is one of the components of the MHV RNA transcription or replication complex, and the crosstalk between hnRNP A1 and another viral or cellular RNA-binding protein (designated X in Figure [8](#cdd462f8){ref-type="fig"}) is essential for MHV replication and transcription. The X protein binds to the C-terminus of hnRNP A1 and cooperates with hnRNP A1 to recruit more proteins to form the transcription or replication complex. The C-terminal-deletion mutant of hnRNP A1 loses the ability to interact with the X protein and to bring it into the initiation complex, resulting in an inhibition of MHV RNA transcription and replication. The residual replication and transcription activities of MHV RNA in the absence of functional hnRNP A1 may be due to a limited affinity of the X protein to a *cis*-acting signal that is only present in MHV genomic RNA (site B). On the other hand, DI RNAs may lack this *cis*-acting signal. When the crosstalk between the X protein and hnRNP A1 is abolished by the dominant-negative mutant of hnRNP A1, the X protein can no longer participate in the formation of the initiation complex, resulting in a complete loss of DI RNA replication.
In summary, our data provide direct experimental evidence that hnRNP A1 is involved directly or indirectly in MHV RNA synthesis, probably by participating in the formation of an RNA transcription/replication complex. This finding reveals a novel cytoplasmic function for hnRNP A1.
Materials and methods
=====================
Cells and viruses
-----------------
DBT cells, a mouse astrocytoma cell line ([@cdd462c18]), were cultured in Eagle's minimal essential medium (MEM) supplemented with 7% newborn calf serum (NCS) and 10% tryptone phosphate broth. MHV strain A59 ([@cdd462c38]) was propagated in DBT cells and maintained in virus growth medium containing 1% NCS.
Plasmid construction and establishment of DBT stable cell lines
---------------------------------------------------------------
The cDNA of the murine hnRNP A1 gene was amplified by RT--PCR using RNA extracted from DBT cells and a set of primers representing the 5′- and 3′-ends of hnRNP A1-coding region, and cloned into pcDNA3.1 (Invitrogen, Carlsbad, CA). The 8 amino acid Flag tag was attached to the N-terminus of hnRNP A1 by including the Flag tag in the forward PCR primer. The truncated hnRNP A1ΔC was similarly constructed using a PCR-amplified fragment that represents hnRNP A1 (aa 1--245).
For the establishment of permanent DBT cell lines, pcDNA3.1 alone or the plasmid containing the Flag-tagged hnRNP A1 or hnRNP A1ΔC was transfected into 60% confluent DBT cells using DOTAP according to the manufacturer's instructions (Boehringer Mannheim, Indianapolis, IN). After 4 h, the transfected cells were selected in DBT cell medium containing 0.5 mg/ml Geneticin (G418) (Omega Scientific, Tarzana, CA) for 10 days. Single colonies were then collected and cultured individually for 10 additional days before screening for the expression of Flag-tagged proteins.
Antibodies
----------
The polyclonal rabbit antibody against p22 was a gift from Dr Susan C.Baker at Loyola University, IL. The chicken polyclonal antibody against hnRNP A1 was produced by Aves Labs, Inc. (Tigard, OR) by immunizing chickens with the purified mouse hnRNP A1 protein expressed in bacteria. The polyclonal anti-Flag antibody was purchased from Affinity Bioreagents (Golden, CO). The goat polyclonal antibody against actin was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). The mouse monoclonal antibody against the N protein has been described previously ([@cdd462c10]).
Examination of growth rate of permanent DBT cells
-------------------------------------------------
Equal numbers (1 × 10^5^) of DBT-VEC, DBT-A1 and DBT-A1ΔC cells were plated in 10-cm culture plates and maintained in culture medium for 4 days. Cells were trypsinized, stained with Trypan Blue (Gibco-BRL, Grand Island, NY) and counted at 24-h intervals with a hemacytometer (Hausser Scientific, Horsham, PA).
Plaque assay
------------
DBT cells in 10-cm plates were infected with MHV-A59 at an m.o.i. of 2. After 1 h for virus adsorption, the cells were washed three times with serum-free MEM, which was then replaced with virus growth medium containing 1% serum. At 1, 6, 8, 10, 14 and 24 h p.i., 1 ml of medium was taken from each plate for plaque assay.
\[^3^H\]uridine labeling of MHV RNA
-----------------------------------
Cells plated in 6-well plates were infected with MHV-A59 at an m.o.i. of 2. At 1 h p.i., 5 µg/ml actinomycin D was added to the virus growth medium to inhibit cellular RNA synthesis. To label newly synthesized MHV RNA, 100 µCi/ml of \[^3^H\]uridine (NEN, Boston, MA) were added to the medium at hourly intervals. After 1 h of labeling, the cells were washed twice in ice-cold PBS and scraped off the plates in 1 ml of PBS. The cells were then collected by centrifugation and incubated in 200 µl of NTE buffer (150 mM NaCl, 50 mM Tris pH 7.5, 1 mM EDTA) containing 0.5% NP-40, 0.5 mM dithiothreitol (DTT) and 400 U/ml of RNasin on ice for 15 min. After centrifugation, 5 µl of the cytoplasmic extract were spotted on a piece of 3 mm paper and incubated with 5% trichloroacetic acid (TCA). The radioactivity remaining on the 3 mm paper was measured in a scintillation counter.
Northern blot analysis
----------------------
DBT cells were infected with MHV-A59 at an m.o.i. of 2. At 8, 16 and 24 h p.i., cytoplasmic extract was prepared as described above and subjected to phenol/chloroform extraction and ethanol precipitation to purify cytoplasmic RNA. Approximately 10 µg of RNA were separated by electrophoresis on a 1.2% formaldehyde-containing agarose gel and transferred to a nitrocellulose membrane. For a better resolution of the DIssE RNA (Figure [6](#cdd462f6){ref-type="fig"}B), RNA was glyoxalated before being electrophoresed on a 1% agarose gel. An *in vitro* transcribed, ^32^P-labeled negative-strand mRNA 7 of MHV-JHM was used as a probe to detect MHV genomic and subgenomic RNAs. For detecting DI RNA species, RNA blots were probed with an RNA representing a sequence complementary to the sequence of the 5′-untranslated region of MHV-JHM RNA, but excluding the leader sequence.
Western blot analysis
---------------------
DBT cells in 6-well plates were infected with MHV-A59 and cytoplasmic extracts were prepared as described previously ([@cdd462c25]) at various time points p.i. The extracts were electrophoresed on a 12% polyacrylamide gel and transferred to a nitrocellulose membrane for western blotting.
Immunofluorescence staining
---------------------------
Cells were washed in phosphate-buffered saline (PBS) and fixed in 4% formaldehyde for 20 min at room temperature, followed by 5 min in --20°C acetone. Primary antibodies were diluted in 5% bovine serum albumin and incubated with cells for 1 h at room temperature. After three washes in PBS, fluorescein-conjugated secondary antibodies were added to cells at 1:200 dilution for 1 h at room temperature. FITC- or TRITC-conjugated secondary antibodies were used to generate green or red fluorescence. Cells were then washed in PBS and mounted in Vectashield (Vector Laboratories, Burlingame, CA).
UV-crosslinking assay
---------------------
UV-crosslinking assay was performed as described previously ([@cdd462c19]). In brief, DBT cell extracts (30 µg protein), 200 µg/ml tRNA and 10^4^ c.p.m. of an *in vitro* transcribed, ^32^P-labeled negative-strand MHV 5′-end RNA (182 bp) were incubated for 10 min at 30°C. Increasing amounts of purified GST (0, 0.5, 1.5 and 5 ng) or recombinant GST--hnRNP A1 fusion protein (0, 1, 3 and 10 ng) were included in the reaction to compete with the endogenous hnRNP A1 for binding. The reaction mixture was placed on ice and UV-irradiated in a UV Stratalinker 2400 (Stratagene) for 10 min, followed by digestion with 400 µg/ml RNase A for 15 min at 37°C. The protein--RNA complexes were then separated on a 10% SDS--polyacrylamide gel and visualized by autoradiography.
GST pull-down assay
-------------------
GST pull-down was performed as described previously ([@cdd462c44]). In brief, GST--hnRNP A1 fusion proteins on glutathione beads (Pierce, Rockford, IL) were incubated with the *in vitro* translated, ^35^S-labeled N protein in 0.3 ml of GST-binding buffer containing 0.1% NP-40 for 2 h at 4°C. The beads were washed five times with the GST-binding buffer containing 0.3% NP-40. Proteins bound to beads were eluted by boiling in Laemmli buffer for 5 min and separated on a 10% polyacrylamide gel.
\[^35^S\]methionine labeling and immunoprecipitation
----------------------------------------------------
DBT cells were infected with MHV-A59 at an m.o.i. of 2. The cells were incubated with methionine-free medium for 30 min before labeling and were labeled in 100 µCi/ml \[^35^S\]methionine starting at 1.5, 7 or 24 h p.i. After labeling for 2 h at each time point, the cells were harvested for protein extraction as described previously ([@cdd462c25]). The protein extracts were immunoprecipitated with anti-Flag antibody-conjugated beads (Sigma, St Louis, MO) in Tm^10^ buffer (50 mM Tris--HCl pH 7.9, 0.1 M KCl, 12.5 mM MgCl~2~, 1 mM EDTA, 10% glycerol, 1 mM DTT, 0.1% NP-40, 1 mM phenylmethylsulfonyl fluoride) at 4°C for 2 h. The immunoprecipitates were washed and separated on a 4--15% gradient SDS--polyacrylamide gel and visualized by autoradiography.
DI RNA transcription and transfection
-------------------------------------
Plasmid 25CAT was linearized by *Xba*I and *in vitro* transcribed with T7 RNA polymerase to produce the DI RNA ([@cdd462c26]). The DI RNA was transfected into MHV-A59-infected DBT cells using DOTAP as described previously ([@cdd462c19]). In brief, ∼80% confluent DBT cells were infected by MHV-A59 at an m.o.i. of 10. At 1 h p.i., the cells were transfected with 5 µg of *in vitro* transcribed DI RNA and incubated at 37°C for the desired lengths of time. To amplify the DI RNA, viruses (P0) were passaged twice in wt DBT cells to generate P1 and P2 viruses.
CAT assay
---------
Cells were harvested at 8 or 24 h p.i. and lysed by freezing and thawing for three times. After centrifugation at 12 000 r.p.m. for 10 min, the supernatant was used in a CAT assay as described previously ([@cdd462c27]).
Figures and Tables
==================
{#cdd462f1}
######
**Fig. 2.** Kinetics of viral infection in DBT cell lines stably transfected with vector (DBT-VEC), the wt hnRNP A1 (DBT-A1) and the C-terminal-deletion mutant of hnRNP A1 (DBT-A1ΔC). (**A**) Syncytium formation. Cells were infected with MHV-A59 at an m.o.i. of 0.5. Following virus removal at 1 h p.i., the cells were maintained in virus growth medium containing 1% NCS. Pictures were taken at 8 and 14 h p.i. (**B**) Kinetics of virus production. DBT-VEC, DBT-A1 and two clones of DBT-A1ΔC cells were infected with MHV-A59 at an m.o.i. of 2. Supernatant was collected at 1, 6, 8, 10, 14 and 24 h p.i. for plaque assay.
{#cdd462f3}
{#cdd462f4}
![**Fig. 5.** Kinetics of MHV RNA synthesis in DBT cells. (**A**) \[^3^H\]uridine labeling of MHV RNA in DBT cells. DBT-VEC, DBT-A1 and DBT-A1ΔC cells were infected with MHV-A59 at an m.o.i. of 2. At 1 h p.i., serum-free medium was replaced by virus growth medium containing 1% NCS and 5 µg/ml actinomycin D. \[^3^H\]uridine (100 µCi/ml) was added to the infected cells at 2, 3, 4, 5, 6, 7, 8, 9, 16 and 24 h p.i. After 1 h labeling, cytoplasmic extracts were prepared and precipitated with 5% TCA. The TCA-precipitable counts were measured in a scintillation counter. (**B**) Northern blot analysis of MHV genomic and subgenomic RNA synthesis in DBT cells. Cytoplasmic RNA was extracted from MHV-A59-infected cells at 8, 16 and 24 h p.i. for northern blot analysis. The naturally occurring DI RNA of MHV-A59 is indicated by an arrow.](cdd462f5){#cdd462f5}
{#cdd462f6}
![**Fig. 7.** Mechanism of the inhibition of MHV RNA transcription by hnRNP A1ΔC. (**A**) Competition of the binding of endogenous hnRNP A1 to MHV RNA by hnRNP A1ΔC. DBT cytoplasmic extracts (30 µg protein) were incubated with a ^32^P-labeled MHV (--)-leader RNA in an *in vitro* UV-crosslinking assay. Increasing amounts of GST (0, 0.5, 1.5, 5 ng) and GST--hnRNP A1ΔC proteins (0, 1, 3, 10 ng) were added to the reaction mixture to compete for the binding. (**B**) GST pull-down assay of the interaction between hnRNP A1 and the N protein. *Escherichia coli*-expressed GST fusion hnRNP A1 proteins (wt and truncation mutants) were incubated with the *in vitro* translated, ^35^S-labeled N protein. The complexes were pulled down by glutathione beads and analyzed on a 10% polyacrylamide gel. (**C**) Co-immunoprecipitation of the wt and mutant hnRNP A1 with an MHV ORF 1a product, p22. Cytoplasmic protein extracts were prepared from MHV-A59-infected DBT-VEC, DBT-A1 and DBT-A1ΔC cells and immunoprecipitated with anti-Flag antibody-conjugated beads. The immunoprecipitates were subjected to western blotting with a Flag antibody (top) and a rabbit polyclonal antibody against p22 (bottom). (**D**) Interaction of the wt and mutant hnRNP A1 with cellular proteins. DBT-VEC, DBT-A1 and DBT-A1ΔC cells were infected with MHV-A59. At 1.5, 7 and 24 h p.i., 150 µCi/ml of \[^35^S\]methionine were added to the infected cells following 30 min incubation in methionine-free medium. After labeling for 2 h, cytoplasmic protein was extracted and immunoprecipitated with anti-Flag antibody-conjugated beads. The immunoprecipitates were separated on a 4--15% gradient SDS--polyacrylamide gel and autoradiographed.](cdd462f7){#cdd462f7}
{#cdd462f8}
Acknowledgements
================
We thank Dr Susan C.Baker at Loyola University, IL, for generously providing the polyclonal antibody against p22. We also thank Drs Jong-won Oh, Guang Yang, Deborah R.Taylor and Peter Koetters for their helpful discussions. This work was partially supported by a National Institutes of Health research grant. S.T.S. is supported by a postdoctoral fellowship from the National Institute for Allergy and Infectious Diseases, National Institutes of Health. M.M.C.L. is an investigator of the Howard Hughes Medical Institute.
[^1]: Present address: Department of Basic Medical Science, 259 Wen Hwa 1st Road, Chang Gung University, Tao Yuan, Taiwan 333
[^2]: Corresponding author e-mail: <michlai@hsc.usc.edu>
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .183}
{#sp2 .184}
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
The mammalian inner ear has very limited ability to regenerate lost sensory hair cells \[[@B1],[@B2]\]. While previous studies demonstrate that early postnatal cochlea harbors stem/progenitor-like cells and show a limited regenerative/repair capacity \[[@B3]-[@B13]\], and these properties, however, are progressively lost later during the postnatal development of the inner ear \[[@B5]\].
Despite the pluripotent differentiation potential of these cells, recovery does not occur to any significant extent after damage to hair cells in the adult mammalian cochlea \[[@B3]\]. Nevertheless, some reports show the presence of limited numbers of nestin-positive stem cells in adult mouse organ of Corti \[[@B5]\], suggesting that there may be some intrinsic potential for repairing hair cells. Therefore, it would be of great interest to explore whether there is the ability of sphere formation from cells harvested during later postnatal and mature stages.
It is well known that the adult mammalian cochleae lack regenerative ability and the irreversible degeneration of cochlear sensory hair cells leads to permanent hearing loss \[[@B14]\], but little is known about the genes and pathways that are potentially involved in this difference of the regenerative/repair potentialities between early postnatal and adult mammalian cochlear sensory epithelia (SE) \[[@B14]\]. The aims of this study are: 1) To investigate whether some remnants of regenerative ability of quiescent progenitor/stem cells can be isolated from mature cochlea, we compare the sphere-forming capabilities of cultured cells derived from P1 and P60 SE; 2) To explore changes of genes and pathways underlying the stem/progenitor cells maintenance and the capacity of regeneration/repair between P1 and P60 SE-derived cells. We examined the expression of a number of genes that were known as stem markers and early inner ear cell markers. These genes have been shown to be linked into known networks and pathways implicated in the mammalian cochlea. In addition, we investigate whether these cells can be differentiated into hair cells and supporting cells *in vitro* using immunocytochemistry.
Materials and methods
=====================
Animals
-------
P1 and P60 C57/BL6 mouse pups (Slac laboratory animal, Shanghai, China) from different litters were used. Animals were housed with mothers in Animal House (College of Chemistry, Chemical Engineering and Biotechnology, Donghua University, China). During this study, animal care and use were in strict accordance with the animal welfare guidelines of the Helsinki Declaration.
Cell culture procedure
----------------------
Dissociated cell cultures were obtained under aseptic conditions from P1 and P60 mice as previously described \[[@B15]\] (Figure [1](#F1){ref-type="fig"}). In brief, SE sheets were isolated from cochleae in Hanks' buffered salt solution (HBSS, Invitrogen) at 4°C, PH 7.4. Tissues were subjected to 0.125% trypsin in PBS solution (Invitrogen) for 15 min, at 37°C, then blocked by trypsin inhibitor and DNAse I solution (Sigma). After gently mechanical dissociation, the pellets were suspended in DMEM/F12 (Dulbecco's Modified Eagle Medium: Nutrient Mixture F-12) 1:1 Mixture (Invitrogen) supplemented with N2 and B27 supplements (Invitrogen), EGF (20 ng/ml) (R&D Systems), bFGF (10 ng/ml) (Wako, Japan), IGF-1(50 ng/ml) (R&D Systems), ampicillin (50 ng/ml; Sigma) and heparin sulphate (50 ng/ml) (Sigma). The suspension was passed through a 70 μm cell strainer (BD Labware) into 6 well plastic Petri dishes (Greiner). Cell cultures were incubated under 37°C, 5% CO~2~, half of the medium was replaced every 2 days. At day 3, cell suspension was replated in new Petri dishes, the attached cells were abandoned. The suspending otospheres obtained from P1 or P60 organ of Corti were assessed in later experiments. For analysis of cell differentiation, we maintained the attached sphere-derived cells in a humidified incubator in a 5% CO~2~ at 37°C in differentiation medium consisting of DMEM/F12 mixed (1:1) supplemented with N2 and B27 (medium and supplements were from Invitrogen), 10% fetal bovine serum (Invitrogen), and ampicillin (50 ng/ml; Sigma). Half of the medium was replaced every 2 days. The differentiated cells were analyzed by immunofluorescence 7 days after plating.
{#F1}
Cell yield and viability
------------------------
The yield and cell viability were determined by using trypan blue vital staining. Four cochleae were dissected from P1 and P60 mice, respectively. The dissociated organ of corti-derived cells were seeded under suspension culture condition, 100 μl cell suspension of each condition was treated separately with 100 μl of 0.4% trypan blue. Using bright field optics, numbers of stained cells with intact plasmamembranes were determined.
Cell proliferation ability was evaluated by 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT) solution (MTT assay kit, Sigma, USA). Briefly, the dissociated organ of Corti-derived cells were plated at 1000 cells/well in 96 well dishes. After the predetermined time points of incubation, the medium on these samples was removed and 10 μl of 5 mg/ml MTT solution was added and assayed according to the manufacturer's instructions. Optical density of solutions in wells was measured at 570 nm using a photometer (MK3 Multilabel Plate Reader, Thermo, USA).
RT-PCR assay
------------
Total RNA was isolated from P1 or P60 mice SE and SE-derived otospheres respectively by using RNeasy Mini Kits (Qiagen), and a mouse embryonic stem cells (ESc) line, G4-2, was taken as positive control to show stem markers. We used 500 ng of total RNA from each group for reverse transcription (RT) by using Superscript III (Invitrogen). We determined the expression of mRNA of stem markers (*nanog, sox2, oct3/4, klf4, c-myc and nestin*), early inner ear cells markers (*jagged1, pax2, brn3.1, bmp7, myosin 7a* and *p27kip1*). Polymerase chain reaction (PCR) analysis was performed by using the following primers pairs: ***nanog*** (fw: AGG GTC TGC TAC TGA GAT GCT CTG, rv: CAA CCA CTG GTT TTT CTG CCA CCG); ***sox2*** (fw: TAG AGC TAG ACT CCG GGC GAT GA, rv: TTG CCT TAA ACA AGA CCA CGA AA); ***klf4*** (fw: CCA ACT TGA ACA TGC CCG GAC TT, rv: TCT GCT TAA AGG CAT ACT TGG GA); ***oct3/4*** (fw: TCT TTC CAC CAG GCC CCC GGC TC, rv: TGC GGG CGC ACA TGG GGA GAT CC); ***c-myc*** (fw: TGA CCT AAC TCG AGG AGG AGC TGG AAT C, rv: TTA TGC ACC AGA GTT TCG AAG CTG TTC G); ***nestin*** (fw: GAT CGC TCA GAT CCT GGA AG, rv: AGA GAA GGA TGT TGG GCT GA); ***jagged1*** (fw: GGT CCT GGA TGA CCA GTG TT, rv: GTT CGG TGG TAA GAC CTG GA); ***pax2*** (fw: CCC ACA TTA GAG GAG GTG GA, rv: GAC GCT CAA AGA CTC GAT CC); ***brn3.1*** (fw: GTC TCA GCG ATG TGG AGT CA, rv: TCA TGT TGT TGT GCG ACA GA); ***bmp7*** (fw: TCT TCC ACC CTC GAT ACC AC, rv: GCT GTC CAG CAA GAA GAG GT); ***myosin 7a*** (fw: CAC TGG ACA TGA TTG CCA AC, rv: ATT CCA AAC TGG GTC TCG TG); ***p27kip1*** (fw: ATT GGG TCT CAG GCA AAC TC, rv: TTC TGT TCT GTT GGC CCT TT); ***GAPDH*** (fw: GGG TGT GAA CCA CGA GAA AT, rv: ACA GTC TTC TGG GTG GCA GT). The RT-PCR experiments were independently repeated 3 times with the same total RNA material. Control RCR reactions were performed lacking cDNA (RT-).
Immunochemistry procedure
-------------------------
To characterize the expression of stem and inner ear progenitor cell maker expression, the first propagation otospheres derived from P1 or P60 organs of Corti were selected for immunostaining analysis. The primary antibodies used were as follows, mouse monoclonal antibody for Oct3/4 (1:500, Santa Cruz Biotechnology, Santa Cruz, CA), for Nestin (1:500, BD Biosciences); goat polyclonal antibody for Sox2 (1:200, Santa Cruz Biotechnology); Pax2 (1:200, BD Biosciences) and for Jagged1 (1:200, Santa Cruz Biotechnology).
The differentiative abilities of P1 and P60 otospheres were characterized by immunocytochemistry using inner ear cell specific markers. After fixation in 4% paraformaldehyde, cells were washed with 0.1 M phosphate-buffered saline (PBS, PH 7.4) for 15 min. Nonspecific staining was blocked in PBS containing 0.2% Triton X-100 (Sigma) and 10% horse serum solution for 30 min. Primary antibodies Myosin7a (1:500, Proteus Biosciences, Ramona, CA), p27kip1 (1:200, NeoMarkers, Fremont, CA), and βIII-tubulin (1:500, Santa Cruz Biotechnology, Santa Cruz, CA) were applied overnight in PBS with 10% horse serum and 0.2% Triton X-100 at 4°C. After 3 times wash, FITC (fluorescein isothiocyanate) or TRITC (Tetramethyl Rhodamine) conjugated second antibodies (Invitrogen) were added for 60 min, counterstaining with DAPI (4′, 6-diamidino-2-phenylindole) (Invitrogen) was performed to visualize cell nuclei. Specimens were examined by confocal microscope (Leica). Negative control experiments were performed as above by omitting the primary antibodies.
Statistical analysis
--------------------
Quantitative data are expressed as the mean ± SD. The statistical process was examined by student's unpaired two-tailed *t*-test. Significant differences are indicated by \* (*p* \< 0. 01).
Results
=======
Cell viability and proliferation
--------------------------------
By staining with trypan blue, we observed that the viable cells from P1 organ of Corti 9.6 ± 0.9 (mean ± SD) × 10^4^ were nearly 2-fold higher compared to P60 5.4 ± 1.1 (mean ± SD) × 10^4^. MTT assay had been used to investigate rates of cell proliferation of P1 and P60 organ of Corti-derived cells. We observed that significantly enhanced proliferation of both cells from day 1 to 3, and maximum effect was produced at day 3 (Figure [2](#F2){ref-type="fig"}). P1 cells showed increasing in proliferation at all time points, while enhanced proliferation was not observed in P60 cells, which changed to reducing from day 3 to 5 and kept stably from day 5 to 7.
{#F2}
Otospheres harvested from P1 and P60 organ of Corti
---------------------------------------------------
As shown in Figure [3](#F3){ref-type="fig"}, cells dissociated from P1 organ of Corti formed nice spheres, and they appeared to continue to proliferate during the course of culture at least 1 week (Figure [3](#F3){ref-type="fig"}A, A׳). While cells dissociated from P60 organ of Corti failed to form nice spheres during culture, only very few spheres could be found in several attempts we made (Figure [3](#F3){ref-type="fig"}B, B׳).
{#F3}
As for the otospheres number analysis, significant difference in sphere-forming cells number between P1 and P60 organ of Corti was observed in present study (Table [1](#T1){ref-type="table"}). In contrast to P60 organ of Corti, the neonatal organ of Corti yielded much more sphere-forming cells, and this situation was maintained during the period of study, no any morphological sign of differentiation was observed. During propagation culture, we found otospheres from P1 could be passaged several propagations, but we could not acquire second generation of otospheres from P60.
######
Otosphere numbers harvested from P1 and P60 cochleae
**Age** **Number of otospheres (n = 5)** **Dissected cochleae numbers** **Treatment**
--------- ---------------------------------- -------------------------------- ---------------------------
P1 3406 ± 245\* 4 Suspension culture 7 days
P60 2 ± 1 4 Suspension culture 7 days
Data are mean ± SD. Significant differences are indicated by \*(*p \<* 0.01).
Expression of stem and inner ear progenitor cell markers in SE and SE-derived otospheres harvested from P1 and P60 cochleae
---------------------------------------------------------------------------------------------------------------------------
The obvious reduction of the ability of cochlear organs from P60 mice to generate otospheres raises the question whether this loss of regenerative capacity is reflected in the expression levels of marker genes for stem cells and/or inner ear progenitor cells. We detected the expression of mRNA of stem markers, *nanog, sox2, klf4, nestin,* and early otic cell markers, *pax2, jagged1, brn3.1, bmp7, p27kip1, and myosin7a* (Figure [4](#F4){ref-type="fig"}). *Oct3/4* was found only expressed in ESc but not in otospheres. No *c-myc* expression was detected in this study. Indeed, the mRNA expression of all stem cell markers that we investigated was stably maintained in P1 and P60 SE as well as SE-derived otospheres (Figure [4](#F4){ref-type="fig"})*.* However, we found inner ear developmental/progenitor markers, i.e. *pax2, jagged1, bmp7, brn3.1, p27kip1* and *myosin7a* showed lower expression in P60 SE and SE-derived otospheres compared with P1 (Figure [4](#F4){ref-type="fig"}).
{#F4}
Immunocytochemistry results
---------------------------
Otospheres from P1 and P60 organs of Corti exhibited the expression of stem markers, Sox2 (Figure [5](#F5){ref-type="fig"}A, A׳) and Nestin (Figure [5](#F5){ref-type="fig"}B, B׳) as well as inner ear progenitors, Pax2 (Figure [5](#F5){ref-type="fig"}C, C׳) and Jagged1 (Figure [5](#F5){ref-type="fig"}D, D′), and no expression of Oct3/4 (data were no shown). As for P1 SE-derived otospheres, Sox2 was found 30 ± 3.4 (mean ± SD)% of total otospheres, Nestin was seen in 91 ± 12.4%, Pax2 was 18 ± 10.2% and Jagged1 was found in all otospheres. However, only very few otospheres from P60 organs of Cotri could be obtained.As for cell differentiation, our results showed the staining of protein markers for inner ear cells upon culturing of P1 and P60 otospheres in differentiation medium (Figure [6](#F6){ref-type="fig"}). All markers employed: Myosin7a (Figure [6](#F6){ref-type="fig"}A and A׳) for hair cells, p27kip1 (Figure [6](#F6){ref-type="fig"}B and B׳) for supporting cells, and βIII-tubulin (Figure [6](#F6){ref-type="fig"}C and C׳) for neural cells (myosin7a in cell processes, βIII-tubulin in the plasma membrane and neurite, and nuclear localization for p27kip1). This confirmed the undifferentiated phenotype of the two kinds of otospheres and their commitment to differentiate to different cell types of the inner ear. However, only very few immunostaining positive cells were observed in the differentiated cells from P60 otospheres. We deduced that the lack of demonstration of hair cell, neural cell and supporting cell differentiation from P60 otospheres is most probably due to their limited cell number. The other reason may also be related to the relatively maturation of P60 cochlea compared with P1.
{#F5}
{#F6}
Discussion
==========
Hearing loss is often caused by loss of hair cells (HCs) due to a variety of factors that generate oxidative stress including noise, ototoxic drugs, cisplatin or aging, etc. \[[@B2],[@B16]\]. The primary cause of hearing loss is the damage or death of the sensory cells in the auditory part of the inner ear, i.e. the cochlea. In mammals, HCs are produced only during embryonic development and do not regenerate, if these cells are lost during postnatal and adult life, the result is profound sensorineural deafness \[[@B17]\]. The adult mammalian cochlea lacks regenerative ability and the irreversible degeneration of cochlear sensory hair cells leads to permanent hearing loss \[[@B17]\]. The previous studies show that early postnatal cochleae harbor stem/progenitor-like cells and show a limited regenerative/repair capacity \[[@B4],[@B7]-[@B15],[@B17],[@B18]\], however, these properties are progressively lost later during the postnatal development \[[@B5]\]. Furthermore, the discovery of stem cells in the adult mouse utricular SE spurred great interest to investigate whether other parts of the inner ear also harbor stem cells that can be isolated by *in vitro* culture \[[@B19]\]. In contrast to adult mammalian vestibular SE, postnatal mouse organ of Corti has very limited capabilities to form sphere cells \[[@B5]\]. Little is known about the genes and pathways that are potentially involved in this difference of the regenerative/repair potentialities between early postnatal and adult mammalian cochlear SE.
To investigate whether some remnants of regenerative ability of quiescent progenitor/stem cells can be isolated from mature cochlea, we compare the cell viability as well as sphere-forming capabilities of cultured cells derived from P1 and P60 organs of Corti, respectively. A notable observation was that the cell viability of P1 organ of Corti-derived cells were 2-fold higher than P60, indicating the cell populations have a wide-ranging alterations at different ages. MTT assay demonstrated that P1 cells showed marked enhancement of proliferative activity during expansion culture *in vitro*, which suggested that P1 cells have higher potential for proliferation compared to P60 cells.
As for the sphere-forming capabilities, we only observed very few sphere formation (2 ± 1, n = 5) from four dissociated P60 SE, which can be negligible compared to P1 tissues (3406 ± 245, n = 5), indicating that the stem cells of the auditory system disappear during the mature stage. Our result is consistent with the data of Oshima's report that the total otosphere numbers are indeed decreased from P21 to P42 organ of Corti and spiral ganglion \[[@B5]\]. Nevertheless, no further data explains the mass loss of sphere formation in the adult organ of Corti. We first reason that the observed loss of sphere formation is reflected the cell number loss during mature. However, when we seeded the same number of cells derived from P1 or P60 cochleae, respectively, we still found that the stark loss of capacity for sphere formation in P60 stage. As for the propagation culture, we found otospheres from P1 can be passaged for several propagations and showed high self-renewal ability and viability. Wherein, we could not acquire second propagation of sphere-forming cells from dissociated primary ototspheres at P60. The reason may be that the self-renewal capacity of the limited numbers of stem cells in adult organ of Corti is progressively lost during the postnatal development. We hypothesize that as an alternative to the loss of stem cells might be interpreted as a loss of stem cell features. The hypothesis is further confirmed by the analysis of RT-PCR expression using stem cell and inner ear progenitor cell markers. The diminishing stem cell features at P60 cochlear cells can provide a reasonable explanation for the long known regenerative inability of the maturing cochlea \[[@B20]-[@B24]\]. This view is supported by our observation that the expression of mRNA of inner ear developmental/progenitor became lower from P1 to P60. Future study will perform real-time PCR or western-blotting experiments to complement the quantitative data. We hypothesize that the loss of stem cells or the loss of stem cell features in the adult organ of Corti is accompanied by a substantial reduction at the expression level of inner ear progenitor cell markers. Whereas the sustained expression of stem cell markers between P1 and P60 tissues is an indication of the maintenance of stemness in mammalian SE, but it is still know little about how to stimulate these quiescent stem cells to self-renewal and proliferate in maturing stage \[[@B25]\].
C57/BL6 is most widely used inbred strain in hearing research. The C57/BL6 mice exhibit a high frequency hearing loss by 3--6 months of age that progresses to a profound impairment by 15 months. This strain has a known age associated hearing loss due to cdh23ahl \[[@B26]\]. Because of the possibility of potential influence of this in ours studies, we examined the otospheres from P 60 mice. We have performed the experiments on ICR mice, and found the expression of stem cell and inner ear progenitor cell markers are associated with age (data were not shown). Because of the possibility of strain differences, it would be great interest to examine the cochlear stem cell phenotypes from different strains, i.e. ICR, BALB/c and C57/BL6 mice in future study.
The results of this approach may provide directions for future investigations into the understanding of the known difference in the ability for regeneration/repair between the early postnatal/developing and adult cochleae. In addition, it would be interesting in future studies to remove or reduce the potential mechanisms of repression to active these stem cells \[[@B27]\], and it is worth pursuing the best growth factor combination that potentially leads to increased cell survival, proliferation and differentiation. Other method, such as gene transfer \[[@B28]\] or iPS cell technology \[[@B29],[@B30]\] can be performed to achieve the goal.
Conclusion
==========
Sphere-forming stem cells from the mouse inner ear are an important tool for the development of cellular replacement strategies of damaged inner ears and are a bona fide progenitor cell source for transplantation studies. Dissociated P60 organ of Corti produced very few otospheres *in vitro*, expressing stem markers, such as sox2 and nestin similarly to P1 SE and otospheres. However, inner ear developmental/progenitor cell show lower expression in P60 stage compared with to P1, which may be contribute the reducing sphere-forming ability from P60 SE. And the lower number of cells for P60 otospheres may relate to its lack of differentiation potential *in vitro*, as opposed to the strong differentiation potential observed *in vitro* for P1 otospheres. However, the limited sphere cell number and restricted differentiation potential observed by us at P60 organ of Corti reinforce this organism as potential experimental studies to search for the mechanisms for organ of Corti regeneration in mammalian cochlea.
Competing interests
===================
The authors declare that they have no competing interests.
Authors' contributions
======================
XXL, JX, XLW and LLY carried out the tissue dissection, cell culture and data collection. JX, XLW and LLY carried out the immunoassay and RT-PCR. XXL performed the analysis and interpretation of data. XXL, YYD and MT were involved in drafting part of the manuscript. YZZ contributed the whole study and participated in the design and coordination of this project as well as manuscript writing. All authors reviewed and approved the final manuscript.
Acknowledgments
===============
The work was supported by the Fundamental Research Funds for the Central Universities (Grant No.2232013D3-13) and Donghua University Scientific Research Foundation for Youths (Grant No.13D210502).
| {
"pile_set_name": "PubMed Central"
} |
###### Article summary
Article focus
=============
- Hypertension carries significant risks but is commonly poorly managed and controlled.
- Telehealth technology appears to be a useful strategy for managing chronic conditions; however equipment deployed is often complex and costly.
- Use of a new, simple and low-cost telehealth intervention was evaluated to establish if it can efficiently manage hypertension in a way that suits patients.
Key messages
============
- A simple, interactive telehealth intervention effective and acceptable way of quickly gaining control of blood pressure (BP) in hypertensive individuals---even among those who had been previously difficult to engage.
- Maximal BP reductions using this type of telehealth intervention appear to occur among hypertensive older adults without chronic kidney disease.
- There may be a place for wider utilisation of this technology to assist in the diagnosis of hypertension, monitoring hypertension and remote clinical management, and in those who find it difficult to attend their general practitioner surgery.
Strengths and limitations of this study
=======================================
- As this is a service evaluation, the results obtained accurately reflect actual use of the technology in the clinical setting.
- Aside from BP measurements in intervention patients, no measurements or tests were undertaken in any patient specifically for this service evaluation, therefore there were some missing data.
- Control patients were less hypertensive at baseline than intervention patients and may have been fundamentally different to intervention patients---it appeared that intervention patients were more likely to have been difficult to control in the past.
Introduction {#s1}
============
Hypertension is an asymptomatic but lethal condition, which carries risks of catastrophic cardiac, renal and cerebrovascular events. However, commonly, it is inadequately controlled.[@R1] Dangers of poor-hypertension control do not simply lie with undertreatment. Overtreatment resulting from inadequate follow-up or falsely high blood pressure (BP) readings taken by health professionals in healthcare settings can potentially result in dizziness, falls and fractures. Given its prevalence among the general adult population, the risks of poorly managed hypertension are substantial with potentially avoidable costs to the health service. In their most recent guidelines for hypertension, the National Institute for Health and Clinical Excellence (NICE)[@R2] highlights the need for good control of hypertension to avoid a consequent treatment-resistant state and also advocate home measurements to prevent treatment of 'white coat syndrome' in an otherwise normotensive individual. Therefore, intermittent clinic BP checks alone are now inadequate and new, efficient and effective means of accurately diagnosing and managing hypertension must be sought.
Telehealth is being increasingly used to assist patients and healthcare professionals responsible for their care to monitor medical conditions. A variety of delivery methods are available, from web-based methods to telephone touch pads[@R3] to complex 'pods' that can transmit multiple readings and symptom reports to healthcare settings from patients' homes. Home telehealth interventions have been shown to substantially reduce mortality, hospital (re) admissions, emergency visits, length of hospital stay and costs of care, while improving quality of life for a number of chronic conditions.[@R4; @R5; @R6; @R7; @R8; @R9; @R10; @R11] A meta-analysis by Bray *et al*[@R12] concluded that self-monitoring results in small but significant reductions in BP. Further, McManus *et al*[@R13] suggest that this may be 'a result of better adherence to treatment'. A recent systematic review and meta-analysis found that the benefits of home BP monitoring can be maximised if telemonitoring is used.[@R14] However, equipment for complex telemonitoring can be expensive, needs extensive training for its use, may require a dedicated non-clinical triage service to receive messages and relay to community nursing teams and, in the case of hypertension, may have functions that are surplus to requirements. This has led, in some areas, to costly equipment being left unused. Therefore a simple, accessible and cost-effective approach to home BP monitoring is desirable.
This service evaluation aims to determine the effectiveness of a simple, interactive telehealth system in reducing BP to a normotensive range in poorly controlled hypertensive adults.
Method {#s2}
======
This service evaluation was undertaken in a materially deprived area of the UK with poor health. Local practices were invited to volunteer to take part in this simple telehealth programme and participating practices were briefly trained in the use of the telehealth intervention. Intervention patients were recruited between April and November 2011 to undertake a 3-month programme using Florence. Patients could continue using Florence after this period if they so desired and at the discretion of the practice. Data collection continued for 6 months from recruitment. Final data collection took place at the end of February 2012.
The simple telehealth intervention utilised in this project required patients to use a home, electronic sphygmomanometer and text via a mobile phone. Patients obtained their BP readings and texted the results to a secure server, 'Florence'. Florence is an interactive service, which reminds patients to text their BP readings each day, and sends reminders if a reading is not received within 2 h. It also sends automatic responses to patients regarding any further required actions based upon the BP readings obtained, from sending a repeat reading if the programme detects that an error may have occurred, through advising patients to contact their doctor or practice nurse. This allows for 'closed loop' monitoring with weekly, or more frequent, review by the patient\'s own general practitioner (GP)/nurse who can log into the secure server and access their results. Messages can be sent back to the patient from their primary healthcare team with advice on continuing management. Although Florence had not been used in this context prior to this project, local pilot use for other conditions resulted in positive anecdotal reports from patients and professionals.
One member of the practice team, usually a practice nurse, led the telehealth programme in each practice. They were responsible for explaining the programme to patients and training them in the use of the software and BP recording devices. Patients were invited to be included in the intervention according to two criteria: Patients with chronic kidney disease (CKD) stages 3 or 4 with BP\>130/85 mm Hg.Patients \>50 years without CKD stages 3--5, with recent BP\>140/90 mm Hg despite prescribed antihypertensive medications.
For each intervention patient, three similar control patients were identified, who underwent usual clinical care. Practice staff were asked to match controls according to inclusion criteria and demographic details; then patient records of control patients were used for data collection while they underwent usual care. Intervention patients were invited to undertake the programme for 3 months or for less time if they became normotensive. Patients were able to continue using the system at the discretion of their practice if they wished to after this time.
Baseline data were collected by practice staff using a proforma for all intervention and control patients. This requested information about demographics, medications, comorbidities, BP at recruitment, most recent BP prior to recruitment and BP 1 year before, most recent estimated glomerular filtration rate (eGFR) and healthcare service use over previous 12-month period. This baseline data allowed calculations of change in BP control as well as comparison between the intervention and control patients at baseline and end of the evaluation. Practices were then asked to complete a proforma which asked for information about BP readings, eGFR, medications, comorbidities and healthcare service use every 2 months for intervention patients for a total of 6 months. Due to administrative burden, data collection occurred every 3 months for control patients. No specific tests of eGFR were organised purely for the sake of this evaluation. Therefore, eGFR results were only available if the practice had initiated this test as part of usual care. Qualitative data were collected at various stages throughout the programme using questionnaires, text messages and a discussion group, the results of which are discussed in an accompanying paper.
Data were collated and analysed using PASW Statistics 18 (Release 18.0.0, 2009) and Microsoft Excel. Missing data were coded as such and requests to practices for these data were made to promote as complete a data set as possible. Intervention patients were analysed as such even after they stopped using Florence. BP readings that were physiologically impossible (eg, diastolic BP higher than systolic BP (SBP)) and those that appeared to be typos and duplicates were excluded. Due to incomplete data collection in some cases, the denominator varied for different calculations. Denominator values used are given in tables and figures.
Home BP readings were adjusted up by 5/5 mm Hg, as per current NICE guidance,[@R2] to ensure that comparisons of BP readings between intervention (in whom the predominant place of BP measurement would be at home) and control patients (in whom the predominant place of BP measurement would be at the GP surgery) were fair.
Control of, and change in, BP and eGFR over the 3-month programme period and follow-up for 6 months from baseline was identified and average changes for both intervention and control groups were calculated and compared for significant differences using a paired sample and an independent sample t test, respectively. The two groups were also examined for variations in medication and healthcare service usage and negative hypertension-related events (eg, stroke/transient ischaemic attack (TIA)) and medication-related events (eg fall and fracture). Statistical significance was assumed at the level of p\<0.05.
Results {#s3}
=======
In total, 490 patients were recruited, of whom 126 were intervention patients and 364 were controls. Of the intervention patients, two were excluded from analyses (with their respective controls) as the patients did not submit any BP readings correctly. Four practices did not identify sufficient control patients, for each intervention patients thus there are eight fewer control patients than would be expected according to the number of intervention patients.
At the point of final data collection, the duration of use of Florence was unknown for one patient and 6 months had not yet elapsed since recruitment for five patients (see [figure 1](#BMJOPEN2012001391F1){ref-type="fig"}). Fifty-one patients stopped using Florence 3 months from baseline, as per the intervention protocol. An additional 37 patients continued to use Florence beyond the required 3 months, of whom 19 continued to submit BP readings to Florence for at least 6 months postrecruitment. Of the 35 patients who did not complete 3 months using Florence, 18 stopped using it as their BP had become controlled, 15 patients chose to stop using it and the reason is unknown for 2 patients. Among those patients who chose to stop using Florence before they had completed the 3-month programme, reasons included the patient left the practice (n=2), family and/or social commitments prevented continuation (n=2), abroad for significant period(s) of time (n=2), did not wish to continue (n=2), difficulty using the system (n=1), problems texting (n=1), difficulty relaxing to take BP (n=1), found taking BP anxiety provoking (n=1), spouse very ill (n=1), incorrectly thought the programme period had ended (n=1), preferred to see GP face to face and the patient only wanted to use it short term (n=1).
{#BMJOPEN2012001391F1}
\*3 months defined as using Florence for 85--95 days inclusive.
[Table 1](#BMJOPEN2012001391TB1){ref-type="table"} outlines the baseline data while [figure 2](#BMJOPEN2012001391F2){ref-type="fig"} represents an overview of the patients at baseline and 3-month postrecruitment. The average age of all patients involved in this project was 59.7 years, with intervention patients having an average age of 58.9 years and control patients having an average age of 60 years. There were no statistically significant differences between participating and control patients with regard to their age, most recent eGFR and number of antihypertensive medications prescribed at baseline. There was no significant difference between intervention patients and controls in the comorbidities they had at baseline, except for osteoporosis for which 4% of intervention patients and \<1% controls had been previously diagnosed.
######
Baseline data
Intervention patients Control patients
--------------------------------------------------------------------------------------------- ----------------------- ------------------
Number of practices 10
Number of patients 124\* 364
Recruitment date 14/04/11 to 07/11/11
Average age at baseline 59 years 60 years
Age range 25--86 years 36--87 years
Number of females 49/124=40% 146/361=40%
eGFR 1 year ago 71 76
Most recent eGFR 77 79
On hypertension register and attended for review in last 12 months 95/109=87% 305/348=88%
On CKD register and attended for review in last 12 months 17/22=77% 54/59=92%
Reason for inclusion:
CKD stage 3 or 4 with BP\>130/85 mm Hg 23 (18.5%) 72 (19.8%)
Patient \>50 years and has BP\>140/90 mm Hg despite prescribed antihypertensive medication† 101 (81.5%) 292 (80.2%)
\*Two patients were excluded from all analyses as they had not managed to text in blood pressure readings correctly so all their data regarding blood pressure was unusable.
†Inclusion criteria classed as this if practices did not indicate a clear inclusion criterion or if a diagnosis of hypertension was being sought and patient did not have CKD.
BP, blood pressure; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate.
{#BMJOPEN2012001391F2}
Intervention patients had significantly greater SBP and diastolic BP readings at recruitment and for the year prior to recruitment, see [table 2](#BMJOPEN2012001391TB2){ref-type="table"}. This remains true even when only patients who were hypertensive at recruitment (intervention patients n=84, control patients n=137) were examined, see [table 3](#BMJOPEN2012001391TB3){ref-type="table"} and [figure 3](#BMJOPEN2012001391F3){ref-type="fig"}.
######
BP readings from 1 year prior to baseline to 6-month postrecruitment among all intervention and control patients
------------------- ---------------------------- ---------------------------- ------------------------------- --------------------- -------------------------- ---------------------
**Average BP 1 year ago** **Average recent BP** **Average BP at recruitment**
**Intervention (n=112)** **Control (n=309)** **Intervention (n=110)** **Control (n=240)** **Intervention (n=124)** **Control (n=358)**
Systolic (range) 149\* (115--192) 139 (96--198) 153\* (103--200) 139 (100--190) 146\* (82--194) 136 (87--197)
Diastolic (range) 87\* (54--118) 81 (48--128) 89\* (51--140) 82 (47--116) 86\* (55--115) 80 (40--112)
**Average BP at 1 month** **Average BP at 2 months** **Average BP at 3 months**
**Intervention (n=105)** **Control (n=51)** **Intervention (n=114)** **Control (n=62)** **Intervention (n=102)** **Control (n=67)**
Systolic (range) 140 (111--181) 136 (106--198) 138 (113--173) 140 (99--184) 138 (108--185) 137 (110--175)
Diastolic (range) 86\* (63--129) 80 (49--128) 85 (65--127) 83 (55--113) 86 (64--131) 82 (48--109)
**Average BP at 4 months** **Average BP at 5 months** **Average BP at 6 months**
**Intervention (n=37)** **Control (n=80)** **Intervention (n=33)** **Control (n=68)** **Intervention (n=29)** **Control (n=61)**
Systolic (range) 144 (114--197) 137 (101--175) 137 (101--175) 139 (101--180) 138 (119--186) 135 (100--167)
Diastolic (range) 85 (66--134) 82 (49--112) 83 (62--132) 82 (47--100) 84\* (67--104) 79 (57--100)
------------------- ---------------------------- ---------------------------- ------------------------------- --------------------- -------------------------- ---------------------
\*BP from intervention patients significantly different to control patients at the level of p\<0.05.
######
BP readings from 1 year prior to baseline to 6 months post-recruitment among all intervention and control patients who had systolic hypertension at baseline (SBP\>130 mm Hg if patients had CKD stages 3--4 or SBP\>140 mm Hg if no CKD stages 3--5)
Average BP 1 year ago Average recent BP Average BP at recruitment
------------------- ---------------------------- ---------------------------- ---------------------------- -------------------- ------------------------- --------------------
Systolic (range) 150 (115--192) 146 (111--189) 155\* (108--200) 146 (100--187) 156\* (132--194) 150 (131--197)
Diastolic (range) 87\* (64--118) 83 (48--128) 90\* (51--140) 84 (56--116) 89\* (64--115) 86 (48--112)
**Average BP at 1 month** **Average BP at 2 months** **Average BP at 3 months**
**Intervention (n=74)** **Control (n=20)** **Intervention (n=78)** **Control (n=31)** **Intervention (n=73)** **Control (n=30)**
Systolic (range) 142 (121--181) 143 (110--170) 140\* (118--173) 146 (112--184) 140 (108--185) 141 (118--161)
Diastolic (range) 87 (67--129) 83 (49--108) 86 (65--127) 86 (63--113) 87 (71--131) 83 (53--106)
**Average BP at 4 months** **Average BP at 5 months** **Average BP at 6 months**
**Intervention (n=24)** **Control (n=36)** **Intervention (n=18)** **Control (n=32)** **Intervention (n=18)** **Control (n=17)**
Systolic (range) 146 (114--178) 142 (109--172) 139 (101--175) 145 (101--173) 139 (124--186) 146 (124--167)
Diastolic (range) 87 (72--134) 81 (49--110) 87 (62--132) 82 (47--98) 86 (76--104) 82 (61--92)
\*BP from intervention patients significantly different to control patients at the level of p\<0.05.
BP, blood pressure.
{#BMJOPEN2012001391F3}
Average SBP and diastolic BP readings for all intervention patients fell within the normotensive range from month 1 of the programme (see [table 2](#BMJOPEN2012001391TB2){ref-type="table"}). From subgroup analysis, intervention patients included with uncontrolled hypertension without CKD stages 3--5 became normotensive (≤140/90 mm Hg) by month 2. Despite intervention patients having significantly elevated SBP compared with control patients at baseline and 1 year prior to this, intervention patients had no significant difference in their average SBP from control patients from month 1 of the programme.
Average BP readings from patients with CKD stages 3--4 only fell into the normotensive range (≤130/85 mm Hg) in month 3. Changes in BP from baseline at each month among intervention patients, were greatest and most significant among patients without CKD stages 3--5 who had systolic hypertension at baseline, see [table 4](#BMJOPEN2012001391TB4){ref-type="table"} and [figure 2](#BMJOPEN2012001391F2){ref-type="fig"}. These patients had significant reductions in SBP ranging from −15 to −16 mm Hg during the 3 months of the programme, a significant difference of at least −10 mm Hg continued to be observed up to 6 months postrecruitment (see [table 4](#BMJOPEN2012001391TB4){ref-type="table"} and [figure 4](#BMJOPEN2012001391F4){ref-type="fig"}); controls who were hypertensive at baseline did not have an equivalent reduction in BP identified until month 3.
######
Average change in BP from recruitment to 6-month postrecruitment in intervention and control patients \>50 years included due to having BP\>140/90 mm Hg despite prescribed antihypertensive medications who had systolic hypertension at baseline
Average change in BP from baseline to 1 month Average change in BP from baseline to 2 months Average change in BP from baseline to 3 months
---------------------------- ---------------------------------------------------- ---------------------------------------------------- ---------------------------------------------------- --------------------------------------- ------------------------------------------ -----------------------------------------
Systolic (95% CI) (range) −15.00\* (−18.51 to −11.49) (−57 to +18)† −6.50 (−14.67 to +1.67) (−36 to +27) −16.37\* (−20.42 to −12.33) (−60 to +23)† −7.82 (−15.74 to +0.11) (−38 to +28) −15.88\* (−19.96 to −11.81) (−61 to +18) −11.42\* (−17.96 to −4.87) (−47 to +16)
Diastolic (95% CI) (range) −1.78 (−4.19 to +0.63) (−22 to +19) −4.00 (−8.35 to +0.35) (−21 to +14) −2.89\* (−5.60 to −0.19) (−24 to +19) −4.19\* (−7.68 to −0.69) (−21 to +15) −3.08\* (−5.77 to −0.40) (−23 to +21) −2.67 (−7.18 to +1.84) (−32 to +18)
**Average change in BP from baseline to 4 months** **Average change in BP from baseline to 5 months** **Average change in BP from baseline to 6 months**
**Intervention (n=20)** **Controls (n=26)** **Intervention (n=15)** **Controls (n=22)** **Intervention (n=14)** **Controls (n=12)**
Systolic (95% CI) (range) −10.25\* (−18.53 to −1.97) (−54 to +25) −8.69\* (−13.91 to −3.48) (−37 to +15) −10.87\* (−19.56 to −2.18) (−30 to +18) −3.09 (−10.02 to +3.84) (−33 to +31) −12.57\* (−19.03 to −6.11) (−36 to +6) −2.42 (−13.95 to +9.12) (−40 to +25)
Diastolic (95% CI) (range) +0.45 (−4.85 to +5.75) (−29 to +24) −3.27 (−7.59 to +1.05) (−27 to +18) −1.67 (−7.73 to +4.39) (−15 to +22) −4.09\* (−8.92 to +0.74) (−24 to +17) −3.36 (−7.91 to +1.20) (−15 to +17) −0.17 (−6.28 to +6.62) (−21 to +19)
\*Change in BP from baseline to time point in programme significant at the level of p\<0.05.
†Difference in change in BP between intervention and control patients significant at the level of p\<0.05.
BP, blood pressure.
{#BMJOPEN2012001391F4}
In addition to the changes in BP observed, significantly greater numbers of BP readings were obtained by patients using Florence than those receiving usual care; 13, 12 and 10 readings per month were obtained in months 1--3, respectively, among intervention patients compared with an average of 0.2 readings a month among controls for the same 3 months.
Significantly more changes, 0.31 vs 0.08 (p\<0.05; CI for difference 0.16 to 0.32) were made to the medications of intervention patients, compared with control patients over the 3-month programme, respectively. There was no evidence of inappropriate overprescribing for white-coat hypertension prior to the programme period as no intervention patients had already prescribed antihypertensive medications stopped due to hypotension. However, 25 patients had systolic hypertension at baseline from clinic readings and were found to be normotensive on home readings in months 1--3 of the programme without any change in medication.
No intervention or control patients suffered from stroke/TIA or hip fracture during the 6-month data collection period. Five control patients and no intervention patients experienced falls during the 6-month data collection period; however, this difference was not statistically significant. Three falls were related to icy weather and the fourth was classified as a trip. There was no significant change in healthcare use or eGFR readings between intervention and control patients during the 6-month data collection period.
Discussion {#s4}
==========
This pragmatic service evaluation has demonstrated the use of a simple interactive telehealth intervention in a real-life clinical primary care setting across 10 general practices. Overall, patients using Florence demonstrated a significantly greater average reduction in SBP, compared with their associated controls, during the initial 3-month programme. This was informed by significantly greater numbers of BP readings and may have arisen as a result of significantly more frequent changes in antihypertensive medications enabled by the telehealth system, more appropriate timing of BP recordings, an increased awareness among patients of their BP values and the significance of the same and/or an increased sense of control among patients. A recent qualitative study by Jones *et al*[@R15], examining patients' experiences of using self-monitoring, self-titration and telemonitoring for hypertension, revealed that once patients had been involved in the self-management programme and witnessed the variability of their BP readings they felt that it was inappropriate for changes to be made to medication based on fewer readings.
When results were analysed according to the inclusion criteria the practices allocated the patient to, no significant improvement in BP control was noted among participating patients with CKD stages 3--4 compared with their associated controls. Among the small number of patients who were normotensive at baseline, an elevation in average SBP was noted (see [figure 2](#BMJOPEN2012001391F2){ref-type="fig"}). The greatest average reduction in SBP among participating patients were noted among those patients recruited for uncontrolled hypertension, without CKD stages 3--5 who were hypertensive at recruitment (see [figure 2](#BMJOPEN2012001391F2){ref-type="fig"} and [table 4](#BMJOPEN2012001391TB4){ref-type="table"}).
Due to the protocol of the programme running only for 3 months, the most robust data set relates to this period. After this, data sets become increasingly incomplete as patients left the programme and returned to usual clinical care. Ongoing differences in BP control between intervention and control patients beyond 3 months may thus be masked.
Although the data collection was as complete as possible, robust comparisons are limited in this service-level evaluation, primarily as 'control' patients underwent usual clinical care. This may have involved not having any BP or eGFR readings recorded near to baseline or for the duration of the programme. Further, patients who initially were using Florence but then stopped before the 3-month programme period was complete may not have had any further BP measurements recorded. Therefore there are some missing data, but this has been indicated in each of the figures and tables.
BP readings that were recorded as 'free text', rather than on the practice computer electronic template, may not have been noted during the data collection for control patients. Therefore, if any practices were recording BP values in this way, this could have led to missing data. There is evidence that this may have occurred in some cases, for example, three control patients are recorded as having had hypertension reviews in the healthcare service usage data but no BP recordings were provided for this period. For one of these patients additional antihypertensive medication was started at the hypertension review by the practice but the BP reading is not provided. As the BP readings from Florence were directly forwarded for data input, any poor record keeping among practices will have had a more significant effect on the BP readings of control patients. The resultant effect on these data may be that any excess reduction in BP among intervention patients compared with control patients is underestimated.
Subjective assessment of a 'typo' in BP readings texted to Florence may not have been accurate at all times; however, the occurrence of these was rare and incorrect exclusion of such data is unlikely to have significantly affected the results.
Although prospective data collection was requested and encouraged, due to other pressures on practice staff, retrospective data collection in relation to medication and use of secondary care resources were undertaken for a significant number of cases and controls. This may well have led to inaccuracies in detail about medication changes, for example, as retrospective collection of this data was not always easy. But BP readings for intervention patients were texted in, in real time and readings were printed and forwarded directly from the secure server. BP readings that were recorded appropriately on practice templates for control patients would also have been noted contemporaneously.
Practices did not strictly match the given inclusion criteria, thus resulting in the recruitment of intervention patients at baseline who were subsequently confirmed as not being hypertensive (see [figure 2](#BMJOPEN2012001391F2){ref-type="fig"}). Practices indicated which inclusion criteria they had recruited patients according to and the results were analysed accordingly. Therefore, results of all patients were analysed according to stated inclusion criteria but also a breakdown of patients according to whether they were normotensive or hypertensive at baseline was given. This breakdown means that the outcomes of using the telehealth intervention reported here for all patients demonstrate actual clinical application rather than highly selected cases but also that an overview of a smaller hypertensive sample has been given. Patients who did not meet inclusion criteria according to age were included in all relevant analyses.
A previous systematic review of randomised controlled trials (RCTs) undertaken in both primary and secondary care settings demonstrated that SBP readings in people undertaking home monitoring are significantly lower (4.2 mm Hg, reducing to 2.2 mm Hg when publication bias allowed for) than control patients.[@R16] Although, during this 3-month programme, there was no significant difference in the average SBP among intervention and control patients, the benefit of this intervention should not be overlooked. Anecdotal feedback from practices highlights that intervention patients may have been fundamentally different to control patients. Feedback from staff and patients implied that intervention patients were often those who, due to personal, family, occupational or unknown factors, had been previously difficult to engage or control. This is also indicated by the significantly greater BP readings at baseline and 1 year prior to baseline among intervention patients compared with control patients (see [table 2](#BMJOPEN2012001391TB2){ref-type="table"}). Therefore, although the quick reductions in BP among intervention patients into the normotensive range in line with control patients by month 6 may represent regression to the mean, it is more likely that achieving similarity in BP results represents a greater achievement than the numbers alone report.
McManus *et al*[@R17] reported the results of an RCT in Birmingham, which utilised self-recording of BP at the GP surgery alongside provision of BP targets to patients. This demonstrated similar outcomes to the current service evaluation, that intervention patients had significantly greater reductions in their BP at 6 months from a higher baseline BP. However, they demonstrated that the difference between intervention and control patients was not sustained at 1 year. Therefore, to determine if the benefits of using Florence identified in this evaluation are prolonged, a longer period of follow-up is needed that can also track the use of healthcare resources.
Healthcare professional concerns, detected in a previous UK study investigating the acceptability of the concept of monitoring of BP using mobile phones, which related to inconvenience of 'continuous streams of readings,'[@R18] were addressed by the current programme which was designed to take no longer than 1 min a week per patient for a health professional to review the BP results. Unacceptably high readings were automatically detected by Florence and patients were given appropriate advice on further action via text. Therefore, the concerns detected by Bostock *et al* did not become reality among practices involved in this service evaluation.
Although patients were trained how to take their BP at the start of the programme, no checks were undertaken to ensure patients were measuring BP as per instructions. Therefore, if patients had not taken multiple readings or used a different arm at different times, the recorded BPs may not have been accurate. In terms of this evaluation this may have affected the change in BP observed. However in the long term, a patient missing the difference in BP between their two arms may result in a missed opportunity to detect comorbid or contributory conditions, for example, peripheral vascular disease.[@R19] Therefore safeguards need to be in place during wider implementation to ensure that patients are measuring their BP accurately throughout their time using this system and to ensure that any difference in BP between two arms that does exist is appropriately managed.
This intervention was well received and effective at quickly reducing elevated BP readings to normotensive ranges. The benefit of rolling this type of intervention out more widely to tackle hypertension in the primary care community could be great, provided patients are carefully counselled regarding what it involves and to ensure that they have the physical and cognitive ability to undertake BP readings and send text messages, or that they have regular contact with someone who is willing to do this on their behalf.
To maximise reductions in BP, careful selection of patients onto a simple telehealthcare programme (eg, patients without CKD stages 3--5 who are hypertensive \>140/90 mm Hg) is required. However, anecdotal evidence obtained through the associated qualitative data indicates that less quantifiable benefits can be gained from recruiting patients with uncertain diagnoses of hypertension and/or suspected white-coat hypertension. Although there was no evidence of Florence helping to detect patients who were being over treated due to white-coat hypertension in this case, a systematic review of RCTs has found that home BP monitoring can result in down titration of hypertensive medication compared with usual care.[@R14] Therefore, a broader application of this type of intervention may be considered. Poor previous patient concordance with management of hypertension should not be a barrier for introducing simple telehealth strategies to patients as the flexibility associated with this management strategy may overcome the difficulties that the patients had previously been experiencing in adhering to healthcare advice.
The likely benefits of improved BP control among intervention patients relate to reduced healthcare service use and reduction in BP-associated comorbidity. However, the impact of reducing BP on these factors is likely to be delayed by some months or years while the costs of reducing the BP (using simple telehealth system (system itself and appointments used training patients to use the system) and increased changes to medication) are realised now. Therefore, to undertake a thorough economic evaluation of an intervention such as and to determine the long-term benefits of this strategy requires follow-up of at least a year.
Supplementary Material
======================
###### Author\'s manuscript
###### Reviewer comments
The authors thank John Blackburn, Chris Chambers, James Davies, Mike Dixon, Dave Sanzeri, Diane Wilshaw who work for NHS Stoke on Trent CCG and were part of the programme steering group and/or data collection team.
**Contributor:** EC designed data collection tools, inputted and analysed data and drafted and revised the original paper. RC lead the programme of work and oversaw the processes of data collection, analysis and reporting, she also devised and revised this paper. PO collected data and revised the draft paper. EC, RC and PO are responsible for the overall content as guarantor(s). JB, CC, JD, MD, DS and DW were all involved in running the programme and for data collection.
**Funding:** Funded by Health Foundation.
**Competing interests:** All authors have completed the Unified Competing Interest form at [www.icmje.org/coi_disclosure.pdf](www.icmje.org/coi_disclosure.pdf) (available on request from the corresponding author) and declare that (1) EC, RC and PO have support from NHS Stoke on Trent for the submitted work; (2) EC and RC have no relationships with Mediaburst Limited that might have an interest in the submitted work in the previous 3 years; PO receives royalties from NHS Stoke on Trent for the use of background IP and PO is a minor stock holder of SCL Consultants limited which is the parent company of Mediaburst Limited (3) their spouses, partners, or children have no financial relationships that may be relevant to the submitted work; and (4) EC, RC and PO have no non-financial interests that may be relevant to the submitted work.
**Ethics approval:** This was a service evaluation.
**Provenance and peer review:** Not commissioned; externally peer reviewed.
**Data sharing statement:** There are no additional unpublished data available.
| {
"pile_set_name": "PubMed Central"
} |
Findings
========
*Taenia solium*is a cestode parasite that causes human neurocysticercosis, a public health problem in developing countries. The life cycle includes the adult tapeworm that grows in the human small intestine and cysticerci, the larval stage, which lodge in pig skeletal muscle and brain. The tapeworm has a scolex that anchors in the intestinal mucosa and is followed by a long row of proglottids, the initial ones are smaller and immature, the middle ones are mature and contain sexual organs and the last segments are the biggest and are gravid because they contain around 60,000 eggs each one. Inside the egg a hexacanth embryo, called an oncosphere, is surrounded by an oncospheral membrane and an egg shell or embryophore; proglottids and eggs are liberated with faeces. After ingestion of eggs by swine, the intermediate host, oncospheres are released and liberated from their membrane, becoming activated in the gut in order to cross the intestinal mucosa and transform into cysticerci. When an individual eats insufficiently cooked infected pork meat, the tapeworm develops \[[@B1]\].
Humans can also acquire cysticercosis after ingesting eggs, this phenomenon is associated with poor health education and lack of sanitation; it is prevalent in pork eating countries of Latin America, Asia and Africa, generating neurocysticercosis, the most frequent and devastating parasitic disease of the brain \[[@B2]\]. Vaccines have been developed targeting the oncosphere and preventing establishment of the parasite in immunized pigs. The recombinant *T. solium*oncosphere proteins, designated as TSOL18 and TSOL45-1A, have been found to induce 99.5% and 97.0% protection respectively, in vaccine trials against the experimental challenge of pigs with *T. solium*eggs \[[@B3],[@B4]\]. Thus, it is biologically relevant to identify the presence of these antigens in the developmental stages of *T. solium*.
In order to localize the antigens on the parasite, blood samples from the pigs that were vaccinated with TSOL18, TSOL45-1A or GST (glutathione S-transferase, as carrier protein), were obtained prior and after immunization \[[@B3]\]. Serum was separated and stored at -20ºC until use. Whole immunoglobulin G (IgG) was purified with a commercial kit (Montage Antibody Purification Prosep-A, Millipore, Bedford, MA, USA), the purified IgG was adjusted at 1 mg/ml. Adult *T. solium*worms were obtained in the experimental hamster model \[[@B5],[@B6]\] and the scolex-neck, immature and mature proglottids were separated, included in Tissue-tek (Tissue Freezing Medium, Durham, NC, USA) and frozen at -70ºC. Cysticerci, obtained from a naturally infected pig, were similarly processed. Before use, 4 µm thick sections were obtained, placed on silane-treated slides, air dried for 30 min, fixed in methanol-acetone for 10 min and dried for 15 min at room temperature. Tissues were rehydrated with PBS, permeabilized with 0.2% Triton X-100 in PBS for 15 minutes, treated for antigen retrieval with pronase E for 10 min and blocked with PBS that contained 1% BSA with 10% goat serum for 1 h at room temperature. TSOL18 IgG, TSOL45-1A IgG, GST IgG or pre-immune IgG, diluted 1:100 in the same blocking solution was added to the slide until all the section was covered and incubated overnight at 4ºC. After washing with PBS-0.3%-Tween 20, sections were incubated with fluorescein isothiocyanate (FITC) labelled goat anti-pig IgG (Santa Cruz Biotechnology Inc, Santa Cruz, CA, USA) for 1 hour at 37ºC and, after washing, counterstained with propidium iodine (PI, Sigma, St. Louis, MO).
Gravid proglottids were recovered in different times from two human tapeworm carriers from Honduras, identified as *Taenia solium*by the number of uterine lateral branches and stored at 4ºC in PBS with 1% antibiotic and antimycotic solution for 15-30 days. For use, gravid segments were placed in a sieve, cut with fine sharp scissors and the fragments obtained were further teased to release eggs. The egg solution was centrifuged and washed in PBS several times. Embryophoric blocks were disrupted with 10% sodium hypochlorite by gently mixing them with a glass Pasteur pipette, when most oncospheres were released from the embryophore, they were washed with RPMI-1640 and were activated by incubation in artificial intestinal fluid (1% pancreatin, 0.2% anhydrous sodium bicarbonate and 1% pig bile in RPMI-1640) for 45 min at 37ºC. As described by Kyngdon *et al*. \[[@B7]\], activation was confirmed under the microscope because oncospheres exhibited active and rhythmic movements; they were further centrifuged (1,000 g, 10 min) and washed in fresh RPMI-1640, thrice. Fixation, permeabilization, antigen retrieval, incubation with antibodies and counterstaining was performed as described above but in Eppendorf tubes. After oncospheres were washed, centrifuged and gently mixed, three drops were placed per slide, allowing to air dry at room temperature. Tissue sections as well as embryos were covered with fluorescence mounting medium (Vectashield, Vector laboratories Inc. Burlingame, CA, USA) and a cover slide. Preparations were observed in a Nikon epifluorescence microscope.
Figure [1](#F1){ref-type="fig"} shows the results obtained with tapeworms recovered from the experimental hamster model seen under contrast phase microscopy or with filters for PI or for FITC. The first row corresponds to the section of the scolex-neck and the second row to proglottids. The worm can be clearly seen under phase microscopy, as well as with PI since the tissue was pre-fixed, but minimal reaction was found with FITC, indicating absence of specific staining. Since all samples had the same pattern, irrespectively of the IgG used (anti TSOL18, anti TSOL45, anti GST or pre-immune), only the samples incubated with TSOL18 are shown. Cysticerci were also negative (not shown).
{#F1}
Positive reactions were only detected with anti-TSOL18 and TSOL45-1A IgG on the surface of oncospheres (Figure [2](#F2){ref-type="fig"}) with similar staining in different assays performed. In this case the first row corresponds to oncospheres incubated with anti GST IgG as a negative control; rows 2 and 3 include images obtained with anti TSOL18 and anti TSOL45-1A antibodies, respectively. The columns correspond to contrast phase microscopy, PI, FITC stains and merged images of both dyes. Pre-immune IgG was negative as was GST, therefore it is not shown. For the phase contrast figures, the hooks of the oncospheres (arrows) were set on focus, thus the rest of the organism is relatively out of focus. PI labelling was positive since, after activating the oncospheres, they were fixed as part of the staining process. Specific staining (FITC) was intensely detected only on the surface of the oncospheres.
{#F2}
Three protective antigens, To16, To18 and To45w, were found on the cytoplasm and the secretory granules of bilateral glandular cells of freshly hatched *T. ovis*oncospheres \[[@B8]\], this was further confirmed by quantitative immunogold labelling in penetration glands and secretory blebs \[[@B9]\]. A different staining pattern was found in activated oncospheres that were cultured for 7 days \[[@B8]\], which is similar to the one found in the present study, ie, on the surface of the oncospheres. This location agrees with the demonstration of *in vitro*capability of antibodies to destroy *T. solium*embryos \[[@B7]\]. Furthermore, although the precise time at which taeniid parasites from a challenge infection are killed by immune hosts is not known, *Taenia taeniaeformis*\[[@B10]\]*T. pisiformis*\[[@B11]\] and *Echinococcus granulosus*\[[@B12]\] oncospheres appear to be killed within 24 h after infection. These data indicate the need of exposed antigens in order to trigger an early protective immune response, supporting the importance of the presence of TSOL18 and TSOL45-1A on the surface of freshly hatched oncospheres.
Abbreviations
=============
TSOL18: the *T. solium*18 oncospheral protein; TSOL45-1A: the *T. solium*45-1A oncospheral protein; GST: Glutathione s- transferase; BSA: bovine serum albumin; To16: the *T. ovis*16 oncospheral protein; To18: *T. ovis*18 oncospheral protein; To45w: *T. ovis*To45w oncospheral protein.
Competing interests
===================
The authors declare that they have no competing interests.
Authors\' contributions
=======================
JMO and AF formulated the idea, MRV, PM and RGK participated during experimental process and in the discussion. All authors contributed in writing the manuscript.
Acknowledgements
================
Charles Gauci and Marshall Lightowlers (Veterinary Clinical Centre, Werribee, Victoria, Australia) kindly provided the vaccines. Luis Alonso Herrera and Sae Muñiz for assisted in photomicrographs. The Research and Ethics Committee of the Hospital General \"Dr. Manuel Gea Gonzalez\" approved the protocol and the Internal Committee for Caring and Using Laboratory Animals approved handling of experimental animals.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Given their rapid proliferation and high specificity, phages are extensively used in several applications, including rapid detection of pathogenic bacteria ([@B33]), assessment of water quality and sewage treatment efficiency ([@B10]), and classification and identification of bacterial species ([@B29]). These applications of phages are generally based on the premise that the tail protein can identify receptors on the bacterial surface. Moreover, phage-based disease treatments have attracted recent research attention because of increasing concerns with drug resistance in bacteria ([@B27]; [@B12]; [@B28]). The tail protein of phages is significant for the interaction between bacteriophages and host bacteria, and this interaction is a research topic with considerable potential. [@B36] have shown the application of genetically engineered bacteriophage tail proteins as molecular probes for the sensitive and selective detection of *Salmonella enterica*. [@B26] used the tail protein gp12 of T4 phage *in vivo* in rice to counteract lipopolysaccharide (LPS)-induced inflammation.
Recognition of and connection with host cells are important steps for all phages to infect bacterial cells. In general, phages can only be stably adsorbed onto the host bacterial surface by identifying an exposed structure on the cell surface as the recognition receptor and then injecting their genetic materials into the cytoplasm of the host bacteria ([@B20]; [@B31]). Porins, affinity transporters, and LPS are the most common outer membrane components that act as receptors for tailed phages that infect Gram-negative bacterium ([@B21]; [@B32]). N4 phage is one type of phage that infects the Gram-negative bacterium *Escherichia coli*. The protein gp65, which constitutes a sheath surrounding the tail tube of N4 phage ([@B8]), is adsorbed onto the cell surface by recognition of the outer membrane protein NfrA ([@B16]; [@B17]; [@B25]). [@B25] speculated that direct interaction of NfrA with gp65 is responsible for the initial signaling event leading to N4 DNA and vRNAP injection. *Achromobacter xylosoxidans* A22732 is a Gram-negative bacterium that was isolated from a Chinese patient with pneumonia and is resistant to multiple β-lactam antibiotics, including carbapenems. This dangerous opportunistic pathogen is distributed in the water environment ([@B30]; [@B7]). Previous research identified the N4-like phage phiAxp-3 in hospital wastewater. This phage specifically recognizes and pyrolyzes *A. xylosoxidans* A22732 and therefore has potential for phage therapy. The genome of phiAxp-3 has been fully sequenced and annotated to determine whether there were any potentially "undesirable" genes (e.g., virulent and/or toxic genes) present ([@B22]). The absence of these "undesirable" genes has important safety implications ([@B37]). No toxin, virulence, repressor genes, integrases, recombinases nor any bacterial gene were detected ([@B22]). The genome analyses in previous study provide no evidence of any potentially dangerous transducing ability of this lytic phage, supporting the potential of the phage for biocontrol applications ([@B22]). Experimental results on receptor identification showed that the receptor of phiAxp-3 is LPS ([@B22]). LPS can trigger inflammatory responses in humans and animals. Analysis of the interactions between the tail proteins of phages and the LPS receptor can be used to develop potential drugs for clinical treatment ([@B26]).
In this study, tail sheath protein (TSP) sequences of phiAxp-3 were screened bioinformatically, and a region with the same effect as the full-length TSP of the phage was identified by protein truncation experiments. The characteristics of the truncated protein, including toxicity, were investigated. The results of this study offer theoretical foundations for future applications of the TSP of N4-like phages.
Materials and Methods {#s1}
=====================
Bacterial Strains, Phages, Plasmids, and Media
----------------------------------------------
*Achromobacter xylosoxidans* A22732 and phiAxp-3 were stored in our laboratory and cultured at 37°C. *E. coli* DH5α, *E. coli* BL21 (»DE3) competent cells and protein expression plasmid pET28a were purchased from Beijing BIOMED, Co., Ltd. Luria-Bertani (LB) medium was used for bacterial liquid cultures, and soft-agar medium included an additional 0.7% (wt./vol.) agar.
Cloning of TSP Genes of phiAxp-3
--------------------------------
A 25 μL polymerase chain reaction (PCR) system was established by using phiAxp-3 genomic DNA. **Table [1](#T1){ref-type="table"}** lists the primers used. PCR products were recycled using the QIAquick Gel Extraction Kit. *Nhe*I and *Bam*HI double enzyme digestion was performed on PCR products and carrier plasmids. DNA fragments were ligated overnight at 4°C and the target segments of the TSP gene were cloned into pET28a.
######
Relevant information on clones produced and investigated in this work.
Protein expression Primer F (NheI)/R (BamHI) Location (bp) Additive amount of IPTG (mM/L)
-------------------- ------------------------------------------------------------------------------- --------------- --------------------------------
ORF69 CTA[GCTAGC]{.ul}ATGTCCATCGATGATTATCG/ CGC[GGATCC]{.ul}TTATTCCGGCACAGCCGTAGC 1--4188 1.0
ORF69(595--1395) CTA[GCTAGC]{.ul}GAGGAAGGATGGGGCTGG/ CGC[GGATCC]{.ul}TTATTCCGGCACAGCCGTAGC 1786--4188 1.0
ORF69(795--1395) CTA[GCTAGC]{.ul}GTTGGTTCCAGCCCTGTAA/ CGC[GGATCC]{.ul}TTATTCCGGCACAGCCGTAGC 2386--4188 0.6
ORF69(995--1395) CTA[GCTAGC]{.ul}CCGCCTACGGGTGTG/ CGC[GGATCC]{.ul}TTATTCCGGCACAGCCGTAGC 2986--4188 0.6
ORF69(1195--1395) CTA[GCTAGC]{.ul}AACCCCATGGATGACAT/ CGC[GGATCC]{.ul}TTATTCCGGCACAGCCGTAGC 3586--4188 0.8
ORF69(795--1195) CTA[GCTAGC]{.ul}GTTGGTTCCAGCCCTGTAA/ CGC[GGATCC]{.ul}TTACTCAGGAGGAATCGGTTC 2386--3585 0.8
ORF69(795--1095) CTA[GCTAGC]{.ul}GTTGGTTCCAGCCCTGTAA/ CGC[GGATCC]{.ul}TTAGACCAGATGCTTACCCGTCAC 2386--3285 0.4
Underlining indicates endonuclease restriction sites.
Expression and Purification of phiAxp-3 TSP Variants
----------------------------------------------------
pET28a containing cloned TSP genes from phiAxp-3 was transformed into *E. coli* BL21 (»DE3) cells and then cultured in high-salinity liquid LB medium (10 g/L of NaCl) supplemented with 0.1% ampicillin (37°C, 180 rpm). When the OD~600\ nm~ reached 0.8, IPTG solution (**Table [1](#T1){ref-type="table"}**) was added and then cell was cultured overnight (25°C, 100 rpm).
Bacteria were collected by centrifugation and washed three times with by 0.5× phosphate-buffered saline (PBS). After resuspension in 15 mL lysis buffer, the bacteria were treated by 30% ultrasonic power (Sonics VCX750, United States) to break the cells (2 s on/4 s off) for 10 min. Then, 20 μL of RNase and DNase were added at room temperature for 1 h, and then mixture was centrifuged at 10,000 *g*/min for 40 min at 4°C. The supernatant was collected and purified using a His-tagged-protein Purification Kit (Beijing CWBIO Company) by following the manufacturer's instructions. Proteins before and after purification were analyzed by SDS-PAGE. The purified proteins were quantified with a Protein Quantification Kit (Beijing BIOMED Company).
Blockage of Phage Adsorption by Purified TSP Variants
-----------------------------------------------------
Fresh *A. xylosoxidans* A22732 solution (50 μL, OD~600\ nm~ = 3.0) and 150 μg/mL TSP variants solution (50 μL) were mixed at 180 rpm for 10 min at 25°C. Then, 100 μL of 10^3^ plaque-forming units (PFU)/mL phage were added and mixed at 16,000 rpm for 3 min at 4°C. The phage titer in the supernatant (residual PFU) was determined by counting the bacteriophage plaques on *A. xylosoxidans* A22732. Addition of 100 μL 1× PBS (instead of TSP) was used as the negative control, and addition of 100 μL 10^3^ PFU/mL phage was used as the positive control. The phage titer of the positive control was defined as 100% ([@B18]). The titer was determined using tested phage suspension and agar overlay technique ([@B4]). Each index was tested three times.
Lipopolysaccharide Extraction
-----------------------------
Fresh *A. xylosoxidans* A22732 solution grown overnight at 37°C was centrifuged at 8000 *g*/min for 2 min at 4°C, cells were killed with 0.5% phenol, and the mixture was centrifuged again at 39000 rpm. The bacterial pellet was washed three times with distilled water, lyophilized, treated with 90% phenol/water (1:1), and heated to 65°C. *A. xylosoxidans* A22732 LPS was extracted for 15 min according to previously published methods ([@B39]). The extract was cooled to 4°C, centrifuged at 5000 *g*/min for 30 min, and the aqueous phase was collected. Distilled water was added to the remaining phenol phase and the extraction process was repeated. LPS was purified according to the method of [@B35]. DNase (20 μg/mL) and RNase (40 μg/mL) were added to the extraction and tubes were kept at 37°C for 4 h. Proteinase K (100 μg/mL) was added to the tubes at 65°C for 1 h, and then they were heated at 100°C for 10 min. The tubes were centrifuged at 8500 *g*/min for 15 min. Supernatants were transferred to 15-mL centrifuge tubes, and two volumes of acetone were added to the extracts. Samples were stored at 4°C overnight to precipitate LPS. The samples were then centrifuged at 2000 *g*/min for 10 min at 4°C, and the precipitates were resuspended in 1 mL distilled water. Extensive dialysis against double distilled water at 4°C was performed until residual phenol was eliminated. Finally, the purified *A. xylosoxidans* A22732 LPS was lyophilized, weighed, and stored at 4°C.
Enzyme-Linked Immunosorbent Assay (ELISA) Verification of the Interaction Between LPS and phiAxp-3 TSP Variants
---------------------------------------------------------------------------------------------------------------
The interaction between *A. xylosoxidans* A22732 LPS and TSP variants of phiAxp-3 was tested by ELISA ([@B24]). *A. xylosoxidans* LPS was diluted to 100 ng/mL with buffer containing (per L) 1.6 g Na~2~CO~3~ and 2.9 g NaHCO~3~. Then, 100 μL of LPS solution was immobilized on an ELISA plate and incubated at 37°C for 2 h. The mixture was rinsed with PBS-TW2 (PBS solution with 0.05% Tween 20). Bovine serum albumin (BSA) diluted in PBS-TW2 was then added at a final concentration of 0.1% and the mixture was left overnight at 4°C. The wells were then rinsed twice with PBS-TW2. *E. coli* 0111:B4 LPS was used as a negative control. Different concentrations of His-tagged proteins (1 μg/mL -- 0.12207 pg/mL) were added (100 μL/well) and the plates were incubated for 30 min at 37°C. The unrelated protein His-tagged fucosidase (from *Bifidobacterium longum* XY01) was used as a negative control. The plate was washed five times with PBS-TW2. In this step, proteins that could not bind to LPS were removed. HRP-conjugated anti-6× His antibody (diluted 1/1,000 in PBS-TW2) was added (100 μL/well) and incubated for 20 min at 37°C. The plate was washed again with PBS-TW2 and incubated for 10 min at room temperature with 50 μL/well SigmaFast OPD substrate before the reaction was stopped by addition of 50 μL of 3 M H~2~SO~4~. The optical density (OD) was measured at 450 nm, and each index was measured three times.
Effects of Different pH Values and Temperatures on the TSP of phiAxp-3
----------------------------------------------------------------------
To explore the effects of pH on the TSP, HCl/NaOH were added to PBS buffer to adjust the pH between 4 and 10. TSP variants were pretreated for 10 min at different pHs and different temperatures (4--70°C, in a water bath). Then, the phage adsorption blockage assay was performed to examine the functionality of the treated TSP.
*In Vitro* Cell Toxicity of the phiAxp-3 TSP
--------------------------------------------
The toxicity of the TSP of phiAxp-3 (and variants) was tested using normal human liver cells (LO2; American Type Culture Collection). LO2 cells were cultured in high-glucose complete Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 1% penicillin (100 U/mL), and streptomycin (100 mg/mL) at 37°C in 5% a CO~2~ atmosphere. Subculture was performed every 4 days (until the cell fusion rate reached 80%). One hundred μL of protein solution diluted in DPBS was added in the following amounts: 12, 6, 3, 1.5, 0.75, or 0.375 μg. In the control group, 100 μL DPBS were added. The survival rate of LO2 cells was tested using the 3-(4,5)-dimethylthiahiazo (-z-y1)-3,5-di- phenytetrazoliumromide (MTT) method after 24 h of culture ([@B11]). Each index was measured three times.
Accession Number of Nucleotide Sequence
---------------------------------------
The amino acid sequence of the TSP of phiAxp-3 was obtained from the NCBI database (accession number [ALA45538.1](ALA45538.1)). The accession number of the amino acid sequence of N4 phage gp65 in NCBI is [YP_950543](YP_950543).
Statistical Analyses
--------------------
Nucleotide sequences were compared using DNAMAN software 8.0 (Lynnonon Biosoft, Quebec City, QC, Canada). The mean values of triplicate experimental readings were determined and subjected to ANOVA. The means were separated using Duncan's multiple range test with the aid of SPSS version 19.0 (SPSS, Inc., United States).
Results
=======
Expression and Purification of TSP and Variants of phiAxp-3
-----------------------------------------------------------
The TSP of phiAxp-3 was predicted to be the 69th open reading frame (ORF), a 1395 amino acid protein which shares 54% identity with the amino acid sequence of gp65 of the N4 phage (**Figure [1](#F1){ref-type="fig"}**). To analyze the location(s) of the receptor-binding region of the TSP, *E. coli* BL21(»DE3) cells were transformed with pET28a expression plasmids that expressed His-tagged full-length ORF69, ORF69(595--1395), ORF69(795--1395), ORF69(995--1395), ORF69(1195--1395), ORF69(795--1195), or ORF69(795--1095) (**Figures [2A,B](#F2){ref-type="fig"}**). In this notation, for example, ORF69(595--1395) represents amino acids 595 to 1395 of the phiAxp-3 TSP of ORF69. These proteins were then purified by Ni-affinity chromatography with the expected sizes: 154, 88, 66, 44, 22, 44, and 33 kDa, respectively (**Figure [2C](#F2){ref-type="fig"}**).
{#F1}
{#F2}
Blockage of phiAxp-3 Adsorption by Purified TSP and Variants
------------------------------------------------------------
We hypothesized that the TSP of phiAxp-3 plays a role in receptor recognition. Therefore, it was necessary to test whether the purified TSP and truncations could inhibit phiAxp-3 binding to the host bacterium. The purified proteins were mixed with fresh *A. xylosoxidans* A22732 solution and the plaquing efficiency of phiAxp-3 was measured. TSP, variants ORF69 (i.e., the full-length protein), ORF69(595--1395), ORF69(795--1395), and ORF69(795--1195) could compete with phage phiAxp-3 for adsorption to the host bacteria, thereby inhibiting phage adsorption (**Figure [3](#F3){ref-type="fig"}**), but the other TSP truncations ORF69(995--1395), ORF69(1195--1395) and ORF69(795--1095), could not (*P* \< 0.05). The adsorption assay demonstrated that TSP functions as a receptor-binding protein, and that residues 795 to 1195 of TSP were sufficient to inhibit phiAxp-3 binding to the receptor *in vitro*.
{#F3}
Binding of TSP and Variants to *A. xylosoxidans* LPS *in Vitro*
---------------------------------------------------------------
A previous report demonstrated that LPS is the adsorption target (receptor) of phage phiAxp-3 ([@B22]). The specific interaction of LPS purified from *A. xylosoxidans* with purified TSP and its variants was confirmed by ELISA (**Figure [4](#F4){ref-type="fig"}**). ORF69(595--1395) (**Figure [4B](#F4){ref-type="fig"}**), ORF69(795--1395) (**Figure [4C](#F4){ref-type="fig"}**) and ORF69(795--1195) (**Figure [4F](#F4){ref-type="fig"}**), like the ORF69 (**Figure [4A](#F4){ref-type="fig"}**), interacted specifically with *A. xylosoxidans* LPS and the results revealed a direct correlation between the concentration of each of these TSP variants and the OD in the ELISA. The other three proteins, ORF69(995--1395) (**Figure [4D](#F4){ref-type="fig"}**), ORF69(1195--1395) (**Figure [4E](#F4){ref-type="fig"}**) and ORF69(795--1095) (**Figure [4G](#F4){ref-type="fig"}**), did not interact with *A. xylosoxidans* LPS. Therefore, consistent with the results of the phage adsorption assay (**Figure [3](#F3){ref-type="fig"}**), residues 795 to 1195 of the TSP are the key region for binding to the receptor, i.e., *A. xylosoxidans* LPS.
{#F4}
Effects of Different pHs and Temperatures on the TSP of phiAxp-3
----------------------------------------------------------------
Stability studies of ORF69 and ORF69(795--1195) were conducted at different pH and temperature values. **Figure [5](#F5){ref-type="fig"}** summarizes the results. The optimum pH for the protein function (i.e., binding to the host bacterium) was 7 (**Figure [5A](#F5){ref-type="fig"}**), and the optimum temperatures were between 4 and 25°C (**Figure [5B](#F5){ref-type="fig"}**). The proteins could not effectively compete with phage to adsorb to host bacteria after treatment in acidic (pH 4) or basic (pH 10) conditions.
{#F5}
Safety Testing of TSP on Cell *in Vitro* Cell Cultures
------------------------------------------------------
ORF69 and ORF69(795--1195) were tested on the human immortalized normal hepatocyte cell line LO2 *in vitro* to evaluate their direct toxic effects on cells. Cell viability was assessed by the MTT method. ORF69 was toxic to the cell line when applied with protein doses \>1.5 μg/well in cultures, however, no toxic or antiproliferative effects of ORF69(795--1195) were observed (**Figure [6](#F6){ref-type="fig"}**). This result indicates that the cytotoxic region of the TSP is independent of residues 795 to 1195.
{#F6}
Discussion
==========
Phage phiAxp-3 was isolated from hospital wastewater and specifically lysed *A. xylosoxidans* A22732. phiAxp-3 is an N4-like phage according to morphological study and genome sequencing. However, this novel phage has a different recognition receptor from N4 phage: N4 phage employs the outer membrane protein NfrA as the recognition receptor, whereas phiAxp-3 uses LPS ([@B22]). Gp65 of N4, encoding the TSP, is thought to be the receptor-binding protein. Using the amino acid sequences of TSP of phage N4 as blast query against NCBI database, there is no recognizable sequence similarity between N4 and other reported N4 like phages. This protein was highly unique, the most similar homology searched was the TSP (encoded by ORF69) of phiAxp-3, although they share only 54% amino acid identity and their binding receptors were totally different (outer membrane protein and LPS, respectively). More than 30 N4-like phages have been reported and sequenced, their most notable feature is the giant RNA polymerase carried in the capsid. The tail proteins (such as tail sheath protein, tail fiber protein, or tail spike protein) responsible for receptor binding showed extensive sequence variability among sequenced N4-like phages. Tail proteins of phages are considered to be a major factor in host specificity. This might be one of the reasons why the hosts of N4-like phages are diverse. A previous study showed that 25 N4-like phage genomes share 14 conserved core genes, which do not contain any phage binding genes ([@B5]), the highly unique binding proteins imply a difference in receptor recognition.
In the present study, recombinant full-length TSP (ORF69) of phiAxp-3 and six variants were co-incubated with the host bacterium. The truncated protein ORF69(795--1195) could directly bind to the host bacteria, with the same function as ORF69 (**Figure [3](#F3){ref-type="fig"}**). ELISA confirmed that ORF69(795--1195) could interact directly with *A. xylosoxidans* A22732 LPS *in vitro* (**Figure [4F](#F4){ref-type="fig"}**). [@B3] demonstrated that the T4 bacteriophage tail fiber adhesin for specific LPS recognition is localized near the C-terminus of the protein. In the present study, the essential sequence of phiAxp-3 TSP for specific LPS recognition was also localized near the C-terminus of ORF69. Short proteins recognizing bacterial cells in a specific manner have major implications for better understanding the molecular recognition of cell surfaces and for bacteria detection ([@B3]).
The characterization of potential therapeutic phages or phage preparations should be undertaken. It is necessary to confirm the stability of the therapeutic phages or phage preparations in different physicochemical conditions ([@B13]). Therefore, in this study, stability studies of ORF69 and ORF69(795--1195) were conducted at different pHs and temperatures. Phages may provide good alternatives to conventional antibiotics, especially for treatment of antibiotic-resistant pathogenic bacteria ([@B23]; [@B9]). [@B38] propounded that the therapeutic phages or phage preparations were non-toxic, and other researchers support this view in a recent report ([@B14]).
Almost all phages recognize receptors through the phage tail ([@B6]). Therefore, the characteristics, especially the safety, of tail proteins of phages must be determined. [@B26] demonstrated that the tail protein gp12 of T4 phage is non-toxic. In the present study, the toxicities of ORF69 of phiAxp-3 and ORF69(795--1195) were evaluated using LO2 cells. The survival rate of LO2 cells decreased as the dose of ORF69 (0.375--12 μg/well in cultures) increased, indicating that ORF69 may be toxic to the cells ([@B1]). However, given the same amount of ORF69(795--1195), the survival rate of LO2 cells was \>90% (**Figure [6](#F6){ref-type="fig"}**), indicating that ORF69(795--1195) is non-toxic in this range (doses: 0.375--12 μg/well in cultures).
In this study, ORF69(795--1195) was shown to interact with *A. xylosoxidans* LPS. Recombinant gp12 (the T4 phage tail adhesin) retains the ability to bind LPS, and gp12 was successfully used in an *in vivo* experiment to counteract LPS-induced inflammation in rice ([@B26]). [@B19] demonstrated that a non-toxic LPS binding peptide was an effective anti-inflammatory peptide for the treatment of acute lung injury. [@B2] demonstrated that a synthetic peptide corresponding to amino acid residues 20 through 44 of the neutrophil-derived 37-kDa cationic antimicrobial protein (CAP37 P20--44) can bind lipid A of LPS, which could be useful in attenuating *in vivo* responses induced during endotoxemia, including sepsis. The emergence of antibiotic-resistant bacteria can cause serious clinical and public health problems ([@B15]). N4-like phages have been used as therapeutic agent in phage therapy against *Pseudomonas aeruginosa* infections and have been shown to be safe and effective ([@B34]). Characterization of the TSP of phage phiAxp-3 might expand our knowledge of using phages as an alternative agent to control multidrug-resistant bacteria. As LPS binding proteins/peptides successfully prevented endotoxin-induced responses *in vivo*, ORF69(795--1195) may be used to develop potential drugs for clinical treatment.
Conclusion
==========
In this study, genes for the TSP of the phiAxp-3 and truncated variants were shown to inhibit phiAxp-3 adsorption onto the host bacterial surface. Experimental results showed no significant difference in residual PFU between the experimental groups \[ORF69(595--1395), ORF69(795--1395), and ORF69(795--1195)\] and the control (ORF69). Similar experimental results were observed in the interaction between the TSP variants and LPS from the host bacterium. Therefore, proteins with similar functions to full-length TSP (ORF69) were screened and the shortest was identified: ORF69(795--1195). The *in vitro* cell toxicity of ORF69 and ORF69(795--1195) was tested; ORF69 was highly toxic, but the ORF69(795--1195) was non-toxic when at dose ranges are 0.375--12 μg. These experimental results can provide references for the application of the TSP of phiAxp-3.
Author Contributions
====================
ZZ and XZ designed the research. ZZ and JZ performed the research. JY and AJ contributed the new reagents or analytic tools. XC and RF analyzed the data. XW, HL, WL, and WY provided experimental materials and equipments. ZZ and CT wrote the paper.
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.
**Funding.** This work received support from National Natural Science Foundation of China (Grant No. 31670174).
[^1]: Edited by: *Helene Sanfacon, Agriculture and Agri-Food Canada, Canada*
[^2]: Reviewed by: *Elizabeth Martin Kutter, The Evergreen State College, United States; Maria Rapala-Kozik, Jagiellonian University, Poland*
[^3]: ^†^*These authors have contributed equally to this work.*
[^4]: This article was submitted to Virology, a section of the journal Frontiers in Microbiology
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Nitric oxide (NO) functions widely as a transmitter in the central and peripheral nervous systems, exerting its physiological effects by stimulating receptors having intrinsic guanylyl cyclase activity, thereby leading to cGMP accumulation in target cells ([@b17]). cGMP can then engage a number of mechanisms, including cGMP-dependent protein kinases, phosphodiesterases and ion channels. One class of ion channel that is potentially modulated by cGMP is the hyperpolarization-activated cyclic nucleotide-modulated cation (HCN) channel, which generates an inward current \[termed hyperpolarization-activated current (*I*~h~)\] at membrane potentials negative to approximately −50 mV. *I*~h~ helps set the resting membrane potential and influences neuronal excitability, synaptic integration and other properties ([@b16]). HCN channels are formulated from four subunits, of which two (HCN2 and 4) confer the most marked modulation by cyclic nucleotides. The binding of cyclic nucleotides shifts the voltage dependence to more positive potentials, thereby increasing the available *I*~h~ current and speeding up its activation kinetics.
cAMP is commonly regarded as the natural ligand for HCN channels because it has been found to be 10- to 60-fold more potent than cGMP ([@b14]; [@b34]; [@b54]). Nevertheless, there is evidence from native neurones that the NO-cGMP pathway can also engage this target. For example, exogenous NO modified oscillatory activity in thalamocortical relay neurones by acting on HCN channels ([@b41]); in trigeminal motor and mesencephalic neurones, NO and cGMP reversibly depolarized the membrane and reduced the firing threshold by a presumed action on HCN channels ([@b1]; [@b42]); the cGMP derivative 8-bromo-cGMP concentration-dependently shifted the HCN channel activation curve in sensory ganglia in the depolarized direction ([@b27]); and, in the substantia gelatinosa, 8-bromo-cGMP or exogenous NO enhanced *I*~h~ ([@b30]).
Most recently, studies on rat optic nerve found that NO coming from endothelial cells persistently depolarizes axons through cGMP acting on HCN channels ([@b18]). This finding represented the first evidence that endogenous NO can act through HCN channels in the central nervous system. In addition, the channels were found to be similarly sensitive to exogenous cAMP and cGMP derivatives, in agreement with a previous result from sensory ganglia ([@b27]).
The aim of the present work was to test more directly if neuronal HCN channels serve as targets for the NO-cGMP pathway. For this purpose, neurones of the deep cerebellar nuclei (DCN) were selected based on evidence that they possess cAMP-sensitive HCN channels ([@b13]) and that NO synthase and NO-activated guanylyl cyclase are found abundantly there ([@b46]; [@b15]).
Materials and methods
=====================
Tissue preparation
------------------
All procedures were in accordance with regulations of, and approved by, the UK Home Office. Post-natal day 8--13 Sprague-Dawley rats (Charles Rivers, Ltd) were killed by cervical dislocation and decapitation. Juvenile rats were used because, beyond this age range, the greater extent of myelination diminishes both the visibility and viability of the neurones after slicing ([@b19]). The brain was rapidly removed and the cerebellum dissected out, taking care to remove the meninges, and placed in iced (4°C) sucrose-substituted low Na^+^ artificial cerebrospinal fluid containing (in m[m]{.smallcaps}): 250 sucrose, 2.5 KCl, 6 MgCl~2~, 2.0 CaCl~2~, 1.0 NaH~2~PO~4~.H~2~O, 26.2 NaHCO~3~ and 11 glucose, equilibrated with 95% O~2~/5% CO~2~. Coronal slices (200--350 μm thick) were prepared in this solution using a vibroslicer and then maintained in a holding chamber containing artificial cerebrospinal fluid (119 m[m]{.smallcaps} NaCl, 2.5 m[m]{.smallcaps} KCl, 1.3 m[m]{.smallcaps} MgCl~2~, 2.0 m[m]{.smallcaps} CaCl~2~, 1.0 m[m]{.smallcaps} NaH~2~PO~4~.H~2~O, 26.2 m[m]{.smallcaps} NaHCO~3~ and 11 m[m]{.smallcaps} glucose equilibrated with 95% O~2~/5% CO~2~) first at 38°C for 1 h and at room temperature thereafter. The transient incubation at physiological temperature (20°C) was found to improve the yield of viable neurones located near the slice surface.
Immunohistochemistry
--------------------
Cerebellar slices containing the DCN or the whole cerebellum were fixed in ice-cold, freshly depolymerized paraformaldehyde (1%) in 0.1 [m]{.smallcaps} phosphate buffer (pH 7.4) for 2 h. Tissue was cryoprotected using ice-cold sucrose solution (5% for 3 h and then 20% overnight), then quickly frozen under dry ice in optimal cutting temperature embedding medium and sectioned coronally at 10 μm intervals onto chrome alum/gelatin-coated slides.
An immunoperoxidase staining procedure was applied to characterize the locations of neuronal NO synthase (nNOS) and NO receptor-associated guanylyl cyclases. A peroxidase suppressor in a methanol solution was used to inhibit endogenous peroxidase activity; this was applied for 15 min either before or after the primary antibody. The presence of NO-activated guanylyl cyclase was probed using two rabbit polyclonal antibodies. The first antibody was raised against a synthetic peptide corresponding to the rat β~1~ subunit, which is found in all known NO receptor isoforms \[1 : 500; Cayman Chemical, MI, USA, 160897; peptide affinity-purified rabbit polyclonal anti-rat/bovine/human soluble guanylate cyclase β1 subunit raised against a synthetic peptide containing rat cyclase amino acids 188--207; suitable for immunohistochemistry as recommended by [@b15]\]. The second antibody was raised against a synthetic peptide corresponding to the rat α~1~ subunit \[1 : 400; Sigma, Gillingham, UK, G4280; rabbit polyclonal anti-rat/human/murine/bovine soluble guanylate cyclase α~1~ subunit; suitable for immunohistochemistry as recommended by [@b35]\]. nNOS was also located using two antibodies: a rabbit anti-rat nNOS antibody \[1 : 500; Zymed/Invitrogen 61-7000; epitope affinity-purified rabbit polyclonal anti-rat nNOS raised against a recombinant protein consisting of 195 amino acids from the N-terminal of rat nNOS protein; suitable for immunohistochemistry as recommended by [@b18]\] and a sheep anti-rat nNOS antibody \[1 : 10 000; a gift from Dr P. C. Emson, The Babraham Institute, Babraham, UK; suitable for immunohistochemistry as recommended by [@b18]\].
Following rehydration of the slides with 0.1 [m]{.smallcaps} Tris-buffered saline containing 0.1% Triton-X-100, the tissue was blocked with 20% donkey serum and then incubated with primary antibody and 1% donkey serum overnight at 4°C. The slides were rinsed twice with the Tris-buffered saline/Triton-X-100 mixture and then once with Tris-buffered saline, and incubated with donkey anti-rabbit biotinylated secondary antibody for 1 h at room temperature. Slides were washed and incubated with Vectastain elite ABC (Vecto Labs Ltd, Peterborough, UK) for 45 min, stained for 4 min with 0.05% 3,3′-diaminobenzidine (Sigma, Dorset, UK) and then counterstained with Mayer's haemalum for 15 s. Finally, slides were air-dried and mounted in distyrene, plasticizer and xylene medium and viewed using a Leitz Aristoplan microscope. All images presented are representative examples of sections from two or more animals. Brightfield photographs were adjusted using only the 'auto levels' and 'auto contrast' functions in Adobe Photoshop and the panels were then arranged in Adobe Illustrator.
Electrophysiology
-----------------
The slices were allowed to recover for at least 60 min and then one slice was transferred into a submerged slice recording chamber that was perfused continuously with artificial cerebrospinal fluid. All electrophysiological recordings were performed at 30--32°C. To study HCN channel function, 4-aminopyridine (4-AP) (1 m[m]{.smallcaps}) and tetrodotoxin (TTX) (500 n[m]{.smallcaps}) were present in the bath solution throughout; no additional voltage-dependent ion channel blockers were considered necessary because preliminary experiments showed that when CsCl (2 m[m]{.smallcaps}) was subsequently added to block HCN channels little active current remained (see [Fig. 3A](#fig03){ref-type="fig"}). Analysis was performed offline using Clampfit 8.0 (Axon Instruments, Foster City, CA, USA). For all recordings, '*n*' refers to the number of individual cells, usually from different slices.
![Modulation of HCN channels in DCN neurones by cyclic nucleotides in whole-cell recordings. (A) Sample traces showing that 10 μ[m]{.smallcaps} isoprenaline (IsoP) caused a reversible, negative shift in the holding current, whereas 100 μ[m]{.smallcaps} DEA/NO had no obvious effect and the HCN channel blocker CsCl inhibited the slowly activating inward current. The inset shows the voltage-clamp protocol, the holding potential being −43 mV. (B and C) Steady-state activation curves constructed from tail currents in A. (D and E) Summary data of the effect of 10 μ[m]{.smallcaps} isoprenaline and 100 μ[m]{.smallcaps} DEA/NO on *V*~1/2~ (*n* = 4 and 3, respectively; raw data in grey, means ± SEM in black). The shift in the presence of isoprenaline was significant (*t*~3~ = 4.66, *P* = 0.019), whereas the effect of DEA/NO was not (*t*~2~ = 1.19, *P* = 0.36). (F) A series of interleaved experiments showing the extent of modulation of HCN channel function by cyclic nucleotides added to the pipette solution. The data are expressed as the slope of the activation curves (k; top) and the *V*~1/2~ values (bottom). Each point is the mean ± SEM from two to five separate cells in different slices. \*Significantly different from control by Student's unpaired *t*-test (*P* \< 0.03). k: 100 μ[m]{.smallcaps} cAMP, *t*~8~ = 1.05, *P* = 0.32; 100 μ[m]{.smallcaps} cGMP, *t*~7~ = 0.33, *P* = 0.75; 1 m[m]{.smallcaps} cGMP, *t*~6~ = 0.08, *P* = 0.94; 1 m[m]{.smallcaps} 8-bromo cAMP (8-Br-cAMP), *t*~7~ = 0.62, *P* = 0.55; 100 μ[m]{.smallcaps} 8-bromo cGMP (8-Br-cGMP), *t*~5~ = 0.82, *P* = 0.45; 1 m[m]{.smallcaps} Br-cGMP, *t*~8~ = 0.84, *P* = 0.43. *V*~1/2~: 100 μ[m]{.smallcaps} cAMP, *t*~8~ = 3.04, *P* = 0.016; 100 μ[m]{.smallcaps} cGMP, *t*~7~ = 1.65, *P*=0.14; 1 m[m]{.smallcaps} cGMP, *t*~6~ = 2.76, *P* = 0.033; 1 m[m]{.smallcaps} 8-Br-cAMP, *t*~7~ = 5.93, *P* = 0.00058; 100 μ[m]{.smallcaps} 8-Br-cGMP, *t*~5~ = 5.34, *P* = 0.0031; 1 m[m]{.smallcaps} 8-Br-cGMP, *t*~8~ = 6.10, *P* = 0.00029. (G) Representative steady-state activation curves for the conditions in F \[the curve for 8-bromo cAMP (8-Br-cAMP) is omitted for clarity\].](ejn0031-1935-f3){#fig03}
Whole-cell recordings were made from large-diameter (dentate or interposed) DCN neurones, visually identified using normal optics and a × 40 water-immersion objective (N.A .75, Zeiss). Pipette solutions initially contained (in m[m]{.smallcaps}): 150 KMeSO~4~, 10 KCl, 10 HEPES, 4 NaCl, 4 MgATP and 0.4 NaGTP. The pH was adjusted to 7.39--7.4 and the osmolarity to 280--290 mOsm/L. Open pipette resistance was 2--4 MΩ. Under voltage-clamp conditions, currents were recorded using an Axopatch 1D amplifier and pClamp7 (Axon Instruments) in neurones clamped at holding potentials from −40 to −55 mV. Resting membrane potentials (observed before and after TTX application) showed regular cycling between −37 and −68 mV. In all experiments, 80% compensation of the series resistance was applied and junction potentials were not corrected for. The recordings were filtered at 1 kHz and sampled at 2 kHz. The mean input resistance (*R*~input~) of the cells was 63 ± 8.6 MΩ (range 33--92; *n* = 7), similar to previous reports ([@b5]). Patch longevity was usually a maximum of 30 min and, because of this limitation, CsCl (2 m[m]{.smallcaps}) was used to block *I*~h~ in preference to the organic blockers, such as ZD7288, which require a long time to exert their effect.
A series of long (3--4 s) voltage steps (*V*~step~) were used to analyse *I*~h~ ([Fig. 1](#fig01){ref-type="fig"}) and double exponential lines were fitted to the charging curve to give the fast and slow time constants for activation. The voltage dependence of *I*~h~ activation was assessed by measuring the activation and deactivation tail currents upon return to a fixed potential (approximately −85 mV for 2 s; *V*~fixed~ in [Fig. 1](#fig01){ref-type="fig"}) to remove the effect of the driving force; the tail current values were measured as the difference in current at the start and end of this voltage step. These current amplitudes were used to construct steady-state activation curves, which showed a typical S-shaped dependence on voltage and were fitted to the Boltzmann function. where *V* is the membrane potential at *V*~step~, *V*~1/2~ is voltage at which the channels are the half-maximally activated and the slope factor, *k*, serves as an indication of the relation between voltage and fractional activation (see [@b9]). Stably clamped cells often became leaky and unusable if prolonged hyperpolarizing voltage steps were repeatedly applied. Because of these technical limitations, we were unable to collect data at potentials where *I*~h~ was fully activated or deactivated. Therefore, the steady-state fractional activation function was obtained using shorter voltage steps and by normalizing the tail currents as follows where *V*~max~ and *V*~min~ are the voltages corresponding approximately to maximum and zero activation of *I*~h~, respectively.
{#fig01}
On obtaining the whole-cell configuration, some cells demonstrated run-down in *V*~1/2~ values and so all recordings were allowed to stabilize over a 5--10 min period before experimental testing began. In order to minimize the effect of this unavoidable whole-cell phenomenon, paired *t*-tests were performed on control and test data. This means that a lack of observed change may be because variations in baseline *V*~1/2~ values could make it difficult to detect subtle changes.
Test compounds were applied in the perfusate for 3--7 min, before the above step protocol was applied to study the properties of HCN channels. To follow the progress of wash-in (and wash-out), a small (approximately 40 mV), short (typically 500 ms) hyperpolarizing step (such as illustrated in [Fig. 6A and B](#fig06){ref-type="fig"}) from the holding potential was applied every 10 s to activate sufficient HCN channels to monitor changes. *I*~h~ was measured as the difference between the point following the capacitative transient and the plateau prior to the offset of the voltage step.
![Voltage-clamp, perforated-patch recordings of the action of PAPA/NO on *I*~h~ in DCN neurones. (A) Example of the time-course of changes in the amplitude of *I*~h~ upon bath application and wash-out of 10 μ[m]{.smallcaps} PAPA/NO. The break indicates when the protocol was applied for determining steady-state activation (see [Fig. 1](#fig01){ref-type="fig"}) in the presence of PAPA/NO. (B) Sample currents from the experiment in A; the protocol is shown in the inset. (C) Steady-state activation curves showing a reversible depolarizing shift in the voltage dependence of HCN channel activation in the presence of 10 μ[m]{.smallcaps} PAPA/NO. (D) In the same cell as C, 10 μ[m]{.smallcaps} PAPA/NO also caused a reversible speeding of the activation kinetics of *I*~h~ as determined by double exponential fits to the charging curves (see [Fig. 1](#fig01){ref-type="fig"}); legend is as for C. The effect of 50 μ[m]{.smallcaps} (E) and 100 μ[m]{.smallcaps} (F) PAPA/NO on the steady-state activation curves for *I*~h~ in different cells. (G) In the presence of 10 μ[m]{.smallcaps} 1*H*-\[1,2,4\]oxadiazolo\[4,3-a\]quinoxalin-1-one (ODQ) throughout the recording, 50 μ[m]{.smallcaps} PAPA/NO did not appear to affect the voltage dependence of activation for HCN channels in this cell. (H) Summary data showing the significant changes in *V*~1/2~ in the presence of 100 μ[m]{.smallcaps} (*t*~2~ = 4.34, *P* = 0.049) and 50 μ[m]{.smallcaps} (*t*~1~ = 107; *P* = 0.0060; non-directional test) PAPA/NO. In the presence of 10 μ[m]{.smallcaps} ODQ, 50 μ[m]{.smallcaps} PAPA/NO had no effect (*t*~1~ = 0.26, *P* = 0.84; non-directional test).](ejn0031-1935-f6){#fig06}
For sharp electrode recording, borosilicate microelectrodes (100--160 MΩ; Clark Capillaries, Reading, UK) were filled with 2 or 3 [m]{.smallcaps} potassium acetate. Blind recordings were made in either the lateral or interposed DCN. Upon entry into the cell, a strong hyperpolarizing current bias was applied to promote stability. The membrane voltage was filtered at 1 kHz and sampled at 2 kHz using an Axoclamp-2B amplifier (Axon Instruments) in bridge mode. The membrane potential was clamped at around −55 mV, as this potential was sufficiently hyperpolarizing to prevent spontaneous, rhythmic firing. The bridge balance was constantly monitored throughout the experiment and adjusted accordingly. The effect of a range of current injections was determined and the intensity adjusted to allow activation of *I*~h~. A step protocol (illustrated in [Fig. 4](#fig04){ref-type="fig"}) was applied every 10 s; a short (500 ms) and small hyperpolarizing step (insufficient to activate *I*~h~) was applied for the calculation of R~input\ short\ step~ when *I*~h~ is not activated and then a 3750 ms larger hyperpolarizing step to activate *I*~h~. The input resistance during this second step (*R*~input~) was determined using the average potential of the plateau that appeared following the voltage sag.
![Current-clamp recordings using sharp electrodes of the effect of NO on *I*~h~ in DCN neurones. (A) Membrane potential response to hyperpolarizing current injections under control conditions (thick line), in the presence of 100 μ[m]{.smallcaps} DEA/NO (thin line) and after wash of DEA/NO (dashed line); the protocol shown in the inset was delivered at 0.1 Hz. The resting membrane potential was −49 mV; traces are individual sweeps. (B) Magnification of the changes in membrane potential to the smaller, shorter initial current injection showing no apparent voltage sag or change in membrane properties in the presence or absence of NO because the voltage response falls at the foot of the steady-state *I*~h~ activation curve (see [Fig. 1](#fig01){ref-type="fig"}); traces are an average of five sweeps for each condition. (C) Summary of the passive input resistance (*R*~input\ short\ step~) calculated from data such as shown in B. The effect of DEA/NO was not significant (*t*~4~ = 2.32, *P* = 0.081). (D) Magnification of the response to the larger, longer current injection in A showing that 100 μ[m]{.smallcaps} DEA/NO caused a reversible decrease in the depolarizing sag in the membrane potential. Inset: 2 m[m]{.smallcaps} CsCl inhibited the depolarizing sag and caused repeated spiking at the end of the hyperpolarizing step; the scales are the same as in the main panel. (E) The rebound voltage changes taking place on removal of the negative current step in D on a faster time-scale. Time-courses of the effect of 100 μ[m]{.smallcaps} DEA/NO on the membrane potential (F), current being injected to maintain it at around −49 mV, on the amplitude of the depolarizing sag (G) and on the input resistance (H) during the long hyperpolarizing current injection. Summary of the changes in sag amplitude (I) and input resistance (J) in response to DEA/NO. Individual cells are represented by small squares and dotted lines; black circles are the means (± SEM). In I the results from one individual cell are hidden under the mean value. The effect of DEA/NO on both the sag (*t*~4~ = 4.12*, P*=0.015) and input resistance (*t*~4~ = 4.30, *P* = 0.013) was significant. All recordings were performed in the presence of 500 n[m]{.smallcaps} TTX and 1 m[m]{.smallcaps} 4-AP.](ejn0031-1935-f4){#fig04}
Gramicidin- and amphotericin-based perforated-patch recordings were also made from visually-identified large DCN neurones. The pipette solution contained (in m[m]{.smallcaps}): 150 KMeSO~4~, 10 KCl, 10 HEPES, 4 NaCl, 0.1 EGTA and 1 MgCl~2~. The pH was adjusted to 7.39--7.4 and the osmolarity to 280--290 mOsm/L. The open pipette resistance was 2--4 MΩ. A gramicidin stock solution was made in dimethylsulphoxide at 1 g/mL. Gramicidin or amphotericin was added to the pipette solution to a final concentration of 120--200 or 90 μg/mL, respectively, just before use and clarified through a 0.45 or 0.8 μm filter. Stable access resistances were obtained at 20--90 min after forming a gigaohm seal. If there was an abrupt increase in the amplitude of recorded current, signalling that the membrane had ruptured, the experiment was terminated. HCN channel function was investigated as for the whole-cell recordings.
Assay of nitric oxide-activated guanylyl cyclase
------------------------------------------------
Guanylyl cyclase activity was measured using a standard assay buffer as a benchmark ([@b22]). Briefly, guanylyl cyclase purified from bovine lung (Axxora UK Ltd, Nottingham, UK) was incubated at 50 ng/mL in assay or test buffer (all containing 0.05 mg/mL bovine serum albumin), pre-warmed to 37°C, and exposed to 20 μ[m]{.smallcaps} of the NO donor 2-(*N*,*N*-diethylamino)-diazenolate-2-oxide (DEA/NO) for 2 min. Samples were then inactivated and cGMP measured by radioimmunoassay. Three independent trials were carried out in each experiment. The standard assay buffer (pH 7.4) contained 50 m[m]{.smallcaps} Tris, 3 m[m]{.smallcaps} MgCl~2~, 0.1 m[m]{.smallcaps} EGTA, 0.05% bovine serum albumin and 1 m[m]{.smallcaps} GTP. Test solutions are described in [Tables 2](#tbl2){ref-type="table"}, [3](#tbl3){ref-type="table"} and [4](#tbl4){ref-type="table"}. During the optimization of the whole-cell solution, the concentrations of free Mg^2+^and of Mg^2+^complexed with ATP, GTP and EGTA were calculated using the computer program Bound and Determined ([@b12]).
######
A Biochemical investigation of the effect of solution composition on NO-stimulated activity of purified guanylyl cyclase: Experiment 1
Guanylyl cyclase activity (μmol/mg protein/2 min)
------------------------------------------------------------------ ---------------------------------------------------
Assay buffer (3 m[m]{.smallcaps} Mg^2+^, 1 m[m]{.smallcaps} GTP) 16 ± 0.7
Assay buffer + KMeSO~4~ (150 m[m]{.smallcaps}) 15 ± 0.7 (*t*~4~ = 0.82; *P* = 0.46)
Assay buffer + KCl (10 m[m]{.smallcaps}) 17 ± 0.5 (*t*~4~ = −1.71; *P* = 0.16)
Assay buffer + HEPES (10 m[m]{.smallcaps}) 14 ± 0.4 (*t*~4~ = 2.68; *P* = 0.06)
Assay buffer + NaCl (4 m[m]{.smallcaps}) 18 ± 2.8 (*t*~4~ = −0.73; *P* = 0.51)
In Experiment 1, the individual whole-cell solution constituents were added separately to the reference assay buffer (50 m[m]{.smallcaps} Tris, 3 m[m]{.smallcaps} MgCl~2~, 0.1 m[m]{.smallcaps} EGTA and 1 m[m]{.smallcaps} GTP at pH 7.4) and had no significant detrimental effect on total cGMP produced on a 2 min exposure to 20 μ[m]{.smallcaps} DEA/NO at 37°C.
######
A Biochemical investigation of the effect of solution composition on NO-stimulated activity of purified guanylyl cyclase: Experiment 2
Guanylyl cyclase activity (μmol/mg protein/2 min)
---------------------------------------------- ---------------------------------------------------
Assay buffer 19 ± 1.0
Whole-cell solution + 0 m[m]{.smallcaps} ATP 0 ± 0.4 (*t*~4~ = 18.11; *P* = 0.0005)
Whole-cell solution + 1 m[m]{.smallcaps} ATP 1 ± 0.3 (*t*~4~ = 17.74; *P* = 0.0006)
Whole-cell solution + 4 m[m]{.smallcaps} ATP 1 ± 0.4 (*t*~4~ = 16.74; *P* = 0.0007)
In Experiment 2, to try to replicate the whole-cell electrophysiological conditions in DCN neurones, the assay buffer was replaced with whole-cell solution (150 m[m]{.smallcaps} KMeSO~4~, 10 m[m]{.smallcaps} KCl, 10 m[m]{.smallcaps} HEPES, 4 m[m]{.smallcaps} NaCl, 0.4 m[m]{.smallcaps} GTP at pH 7.4) with or without the indicated ATP concentrations. Generation of cGMP was virtually abolished, irrespective of the ATP concentration, compared with the assay buffer.
######
A Biochemical investigation of the effect of solution composition on NO-stimulated activity of purified guanylyl cyclase: Experiment 3
Control (no ATP) +ATP (1 m[m]{.smallcaps})
------------------------------------------------------------------------------ -------------------------------------- --------------------------------------
Assay buffer 21 ± 3.2 --
Whole-cell solution + 0.6 m[m]{.smallcaps} Mg^2+^ + 0.4 m[m]{.smallcaps} GTP 2 ± 0.8 (*t*~4~ = 5.74; *P* = 0.005) 2 ± 0.6 (*t*~4~ = 5.96; *P* = 0.004)
+0.1 m[m]{.smallcaps} EGTA 17 ± 2.0 (*t*~4~ = 1.16; *P* = 0.31) 8 ± 0.8 (*t*~4~ = 4.00; *P* = 0.017)
Whole-cell solution + 3 m[m]{.smallcaps} Mg^2+^ + 1 m[m]{.smallcaps} GTP 7 ± 1.6 (*t*~4~ = 3.94; *P* = 0.017) 5 ± 0.2 (*t*~4~ = 5.10; *P* = 0.007)
+0.1 m[m]{.smallcaps} EGTA 15 ± 1.2 (*t*~4~ = 1.91; *P* = 0.13) 8 ± 0.6 (*t*~4~ = 4.19; *P* = 0.014)
In Experiment 3, the introduction of EGTA and free Mg^2+^ (at least 0.6 m[m]{.smallcaps}) restored guanylyl cyclase activity in the whole-cell solution. Two sets of Mg^2+^and GTP concentrations were tested; the first contained free Mg^2+^levels intermediate to the presumed physiological range ([@b23]) and the second had Mg^2+^and GTP concentrations similar to those in the assay buffer. Please note that the values for GTP are the total, final concentrations. All *P*-values represent a Student's unpaired *t*-test between values for the assay buffer and the test condition; in all cases, *n* = 3.
Materials
---------
Forskolin and 1*H*-\[1,2,4\]oxadiazolo\[4,3-a\]quinoxalin-1-one were bought from Tocris Cookson (Bristol, UK), TTX from Latoxan (Rosans, France), KMeSO~4~ from City Chemical (CT, USA) and distyrene, plasticizer and xylene mounting medium and optimal cutting temperature embedding medium from VWR International (Dorset, UK); Mayer's haemalum was purchased from Raymond A Lamb Ltd (Eastbourne, UK) and peroxidase suppressor from Pierce (Rockford, IL, USA). All other specialist chemicals were obtained from Sigma (Gillingham, UK). All standard reagents were obtained from BDH (VWR International). DEA/NO and (Z)-1-\[*N*-(3-ammoniopropyl)-*N*-(*n*-propyl)-amino\]diazen-1-ium-1,2-diolate (PAPA/NO) were obtained from Axxora UK Ltd. Stock solutions were dissolved in water except for forskolin, which was dissolved in dimethylsulphoxide (final concentration in artificial cerebrospinal fluid = 1%), and the NONOate NO donors, which were made up in 10 m[m]{.smallcaps} NaOH (final dilution at least 1 : 1000 into the bath solution).
Statistics
----------
Student's paired and unpaired *t*-tests were conducted using Origin 6 or 7 (OriginLab Corporation, Northampton, MA, USA). Degrees of freedom were *n*−1 for paired measurements and (*n*~1~−1) + (*n*~2~−1) for independent measurements, and *P*-values \< 0.05 were considered statistically significant. For sag amplitudes and input resistance, the mean of approximately five data points before and during the application of NO was used for statistical comparison. Data are presented as means ± SEM.
Results
=======
Previous evidence had suggested that NO receptor-associated guanylyl cyclases ([@b20]; [@b15]) and nNOS ([@b52]; [@b46]) are both present in the adult rat DCN. Because of technical limitations, our electrophysiological recordings were limited to developing, less myelinated tissue ([@b19]) and so expression of the proteins in the juvenile DCN was first examined. Staining for the common β1-subunit of NO-activated guanylyl cyclases was obvious in all layers of the cerebellar cortex, whereas the white matter tracts located between the cortex and DCN were virtually devoid of staining ([Fig. 2A](#fig02){ref-type="fig"}). In contrast, the DCN showed strong staining in both the neuropil and the cytoplasm of the larger neurones ([Fig. 2A and B](#fig02){ref-type="fig"}). A similar pattern was seen when staining for the α~1~-subunit ([Fig. 2C](#fig02){ref-type="fig"}). Abundant nNOS staining (using the Zymed antibody) was observed in the granule cell and molecular layers of the cortex but was faint or absent in the somata of Purkinje cells ([Fig. 2D](#fig02){ref-type="fig"}). Unlike the cyclase, nNOS immunoreactivity was present in only a small number of cells in the DCN but there was an obvious network of nNOS-positive fibres ramifying throughout the nuclei ([Fig. 2E](#fig02){ref-type="fig"}). This pattern of staining in the DCN was also seen using a different antibody ([Fig. 2F](#fig02){ref-type="fig"}).
{#fig02}
The presence of both components of the NO-cGMP signalling pathway in the DCN at the appropriate post-natal age, together with the presence of cyclic nucleotide-sensitive HCN channels in DCN neurones at a similar age ([@b13]), allowed the hypothesis that NO-evoked cGMP signals modify neuronal function by engaging HCN channels to be tested directly. *I*~h~ is an unusual cationic channel that operates at quite hyperpolarized potentials and therefore, in different neuronal types, can contribute to resting membrane potential, length constant and dendritic integration ([@b45]). In DCN neurones, *I*~h~ is active at potentials negative to approximately −50 mV ([@b13]). To study the modulation of HCN channel function by NO, a series of long hyperpolarizing voltage steps (*V*~step~) to activate *I*~h~ were applied followed by a shorter step to a fixed potential (*V*~fixed~; [Fig. 1](#fig01){ref-type="fig"}). The tail currents generated upon return to *V*~fixed~ were normalized and plotted as a function of *V*~step~; these plots gave an indication of the voltage dependence of activation and the potential, *V*~1/2~, at which the channels were half-maximally activated ([Fig. 3](#fig03){ref-type="fig"}). As an initial control, we applied 10 μ[m]{.smallcaps} isoprenaline, a β-adrenergic receptor agonist that classically acts by raising the levels of cAMP; this resulted in the expected positive shift in the voltage dependence of *I*~h~ activation ([Fig. 3A, B and D](#fig03){ref-type="fig"}). Perfusion of the NO donor DEA/NO (100 μ[m]{.smallcaps}), however, had no significant effect ([Fig. 3A, C and E](#fig03){ref-type="fig"}). The control values for these neurones were −88.8 ± 2.34 mV for *V*~1/2~ and 210 ± 33 and 1099 ± 46 ms for the fast and slow activation time constants (*V*~step~ approximately −110 mV), respectively (*n* = 7). To probe cyclic nucleotide sensitivity more directly, cAMP or cGMP was included in the pipette solution. cAMP (100 μ[m]{.smallcaps}) caused a depolarizing shift of approximately 10 mV, as did 1 m[m]{.smallcaps} (but not 100 μ[m]{.smallcaps}) cGMP ([Fig. 3F and G](#fig03){ref-type="fig"}). The 8-bromo-substituted forms of the nucleotides, which are less prone to hydrolysis by phosphodiesterases ([@b53]), were also effective, with 8-bromo-cGMP being maximally active at a concentration (100 μ[m]{.smallcaps}) at which cGMP itself was inactive ([Fig. 3F and G](#fig03){ref-type="fig"}).
One explanation for cGMP but not NO being able to modulate the HCN channels could be that dialysis with the intracellular recording solution inhibits NO-evoked cGMP formation. Indeed, studies in this laboratory have found that the ATP concentration typically used in whole-cell recording solutions (4 m[m]{.smallcaps}) inhibits NO-activated guanylyl cyclase activity by about 90% ([@b47]). Physiological ATP concentrations, by contrast, are around 1 m[m]{.smallcaps} ([@b21]), which causes much less (about 50%) inhibition ([@b48]). With 1 m[m]{.smallcaps} ATP in the whole-cell solution, 100 μ[m]{.smallcaps} (*n* = 2 of 2), 10 μ[m]{.smallcaps} (*n* = 4 of 6) or 1 μ[m]{.smallcaps} (*n* = 5 of 5) DEA/NO had no effect on HCN channels. The *V*~1/2~ values in the latter two cases were −96 ± 4.4 (control) and −98 ± 3.7 mV (10 μ[m]{.smallcaps} DEA/NO; *t*~3~ = 0.81, *P*=0.48), and −87 ± 2.2 (control) and −86 ± 2.4 mV (1 μ[m]{.smallcaps} DEA/NO; *t*~4~ = 0.82, *P*=0.46). In some of these experiments, when patch integrity allowed, 1 m[m]{.smallcaps} 8-Br-cGMP (*n* = 2) or 1 μ[m]{.smallcaps} forskolin (*n* = 1) was subsequently added by perfusion and an approximately 10 mV depolarizing shift in *V*~1/2~ was observed each time (data not shown). Moreover, it was notable that, in two out of the six recordings performed, 10 μ[m]{.smallcaps} DEA/NO caused a depolarizing shift of similar magnitude that appeared to reverse on wash-out. The control *V*~1/2~ values were −88.6 and −93.5 mV, those in the presence of 10 μ[m]{.smallcaps} DEA/NO were −84.6 and −80.3 mV, and after wash-out the values were −88.1 and −89.5 mV, respectively. Inspection of some of the properties of *I*~h~ in all cells exposed to 10 μ[m]{.smallcaps} DEA/NO showed that there was no obvious difference in the resting membrane potential, *V*~1/2~, activation time constants or amplitude of *I*~h~ in those cells that responded compared with those that did not ([Table 1](#tbl1){ref-type="table"}).
######
Comparison of some of the cellular and HCN channel properties for DCN neurones in which *I*~h~ was (cells 1 and 2) and was not (cells A--D) affected by 10 μ[m]{.smallcaps} DEA/NO
Time constants of activation (ms)
--- ----- ------ ----------------------------------- ----- -------
1 −44 −94 917 272 −0.19
2 −43 −89 1104 551 −0.18
A −58 −90 1371 256 −0.30
B −47 −90 1261 198 −0.17
C −47 −98 1604 337 −0.15
D −57 −108 1440 250 −0.21
The resting membrane potential (RMP) was measured at the beginning of each experiment. *V*~1/2~, the voltage at which the HCN channels were half-maximally activated, the amplitude of *I*~h~ and the activation time constants (when stepping to approximately −100 mV) were measured as described in Materials and methods and [Fig. 1](#fig01){ref-type="fig"}; all values were from control recordings prior to exposure to DEA/NO.
The variable response to NO with 1 m[m]{.smallcaps} ATP in the recording solution suggested that other ingredients of the electrode solution may also be detrimental. Accordingly, all components were examined systematically in a biochemical assay of cGMP formation by purified guanylyl cyclase ([Tables 2](#tbl2){ref-type="table"}, [3](#tbl3){ref-type="table"} and [4](#tbl4){ref-type="table"}). The original whole-cell solution was found to be highly inhibitory to NO-evoked activity. Two factors were responsible: lack of EGTA, indicating an inhibitory effect of (background) Ca^2+^ ([@b29]), and free Mg^2+^, which is needed as a cofactor ([@b31]). The final 'physiological' whole-cell solution comprised (in m[m]{.smallcaps}): 150 KMeSO~4~, 10 KCl, 10 HEPES, 4 NaCl, 1 MgCl~2~, 0.1 EGTA, 1 MgATP and 0.4 NaGTP at pH 7.4 and 285 ± 5 mOsm/L. Despite success in the biochemical assay, the new solution failed to confer sensitivity of *I*~h~ in DCN neurones to exogenous NO, when the donor was either DEA/NO (50 μ[m]{.smallcaps}; *n* = 2; *t*~1~ = 0.58, *P* = 0.67; non-directional test; data not shown) or a NONOate with slower release kinetics, PAPA/NO (100 μ[m]{.smallcaps}; *t*~2~ = 1.01, *P* = 0.42; data not shown).
The possibility then arose that, for other reasons, the whole-cell recording technique disrupted NO signal transduction through the cGMP-HCN pathway in most of the neurones. This possibility was first tested using the less disruptive sharp microelectrode recording technique. Under current-clamp conditions, there was a depolarizing sag in the membrane voltage during a long, large hyperpolarizing step ([Fig. 4A, D and I](#fig04){ref-type="fig"}). After cessation of the hyperpolarizing step, there was a prominent rebound depolarization with one or more associated spikes ([Fig. 4A, D and E](#fig04){ref-type="fig"}) presumably caused by the activation of Ca^2+^channels ([@b3]) as TTX-sensitive Na^+^ channels were blocked. The addition of 2 m[m]{.smallcaps} Cs^+^ to block *I*~h~ removed the depolarizing sag ([Fig. 4D](#fig04){ref-type="fig"}, inset). Perfusion of 100 μ[m]{.smallcaps} DEA/NO resulted in a decrease in the sag ([Fig. 4A, D, G and I](#fig04){ref-type="fig"}). The size of the depolarizing sag is an index of the degree of activation of *I*~h~, a decreased size implying that fewer HCN channels are available to be activated by the hyperpolarizing step because more were already active at the holding potential. During the application of NO, there was also a decrease in input resistance during this long hyperpolarizing current injection (*R*~input~; [Fig. 4H and J](#fig04){ref-type="fig"}), as expected from increased HCN channel activation. This interpretation was supported by a lack of NO-induced change in input resistance during a shorter, smaller step current when *I*~h~ is not activated (*R*~input\ short\ step~; [Fig. 4A--C](#fig04){ref-type="fig"}).
There was also a change in the rebound depolarization in the presence of 100 μ[m]{.smallcaps} DEA/NO ([Figs 4A, D, E, and 5A, i--v](#fig04){ref-type="fig"}). The first rebound spike of the depolarization showed a small, non-significant, decrease in amplitude and concomitant non-significant increase in latency to peak ([Fig. 5B, i and ii)](#fig05){ref-type="fig"}. However, the second rebound spike was significantly smaller with a concomitant significant increase in latency to peak in the presence of NO (where it could be resolved; [Fig. 5C, i and ii)](#fig05){ref-type="fig"}.
![Sharp electrode, current-clamp recordings of the effect of NO on rebound potentials in DCN neurones. (A) The rebound potentials of all cells recorded (individually labelled as i--v). The plateau following the depolarizing sag was used as a baseline; scale bars are 100 ms and 10 mV. (B and C) Quantitative data for the rebound potentials in A. Values for the peak amplitude (Bi and Ci) and latency to peak (Bii and Cii) of the first (B) and second (C) rebound spikes are normalized to the mean values during the 300 s control period prior to addition of DEA/NO. Analysis by paired Student's *t*-test of values at *t* = 0 and *t* = 490 s showed that DEA/NO caused non-significant changes in the amplitude and latency of the first spike (*t*~4~ = 2.02, *P*=0.11 and *t*~3~ = 1.32, *P* = 0.28, respectively) but significant changes in these parameters of the second spike (*t*~3~ = 8.46, *P* = 0.0035 and *t*~3~ = 2.91, *P* = 0.026, respectively). All recordings were performed in the presence of 500 n[m]{.smallcaps} TTX and 1 m[m]{.smallcaps} 4-AP.](ejn0031-1935-f5){#fig05}
The positive results using sharp electrodes supported the hypothesis that normal whole-cell recording disrupts the signalling pathway. Accordingly, the next approach was to try the perforated-patch recording technique, which is much less invasive. Successful recordings over a sufficient period with this method, however, proved difficult. Nevertheless, using this method, NO (delivered using 10 μ[m]{.smallcaps} PAPA/NO) was found to cause a reversible increase in the amplitude of *I*~h~ activated by a small hyperpolarizing step ([Fig. 6A and B](#fig06){ref-type="fig"}). This effect was attributable to a depolarizing shift in the voltage dependence of HCN channel activation ([Fig. 6C](#fig06){ref-type="fig"}). The rate of activation was also increased ([Fig. 6D](#fig06){ref-type="fig"}). In all cells (*n* = 6) exposed to PAPA/NO (10--100 μ[m]{.smallcaps}) there was an apparent reversible enhancement of HCN channel function. In addition, the effect appeared to be NO concentration dependent in that cells exposed to 100 μ[m]{.smallcaps} PAPA/NO (*n* = 3) showed a significant depolarizing shift in *V*~1/2~ of approximately 10 mV \[[Fig. 6F and H](#fig06){ref-type="fig"}; control, −89.0 ± 6.0 mV vs PAPA/NO, −78.6 ± 3.9 mV\], 50 μ[m]{.smallcaps} PAPA/NO (*n* = 2) caused a shift of about 5 mV \[[Fig. 6E and H](#fig06){ref-type="fig"}; control, −83.7 ± 4.7 mV vs PAPA/NO, −78.4 ± 4.8 mV\] and with 10 μ[m]{.smallcaps} PAPA/NO (*n* = 1) there was a 4 mV shift ([Fig. 6C](#fig06){ref-type="fig"}). To test if the observed effects of NO were via cGMP, 1*H*-\[1,2,4\]oxadiazolo\[4,3-a\]quinoxalin-1-one, the standard blocker of guanylyl cyclase-coupled NO receptors, was used. In two cells (out of a total of approximately 20 attempts) in which the experiment could be performed, PAPA/NO (50 μ[m]{.smallcaps}) had no observable effect on the steady-state activation curve for *I*~h~ in the presence of 10 μ[m]{.smallcaps} 1*H*-\[1,2,4\]oxadiazolo\[4,3-a\]quinoxalin-1-one \[[Fig. 6G and H](#fig06){ref-type="fig"}; control, −80.9 ± 5.2 mV vs PAPA/NO, −81.5 ± 2.7 mV\].
Discussion
==========
Given the evidence that endogenous NO regulates axonal function in the optic nerve through cGMP and a subsequent activation of HCN channels ([@b18]), the present investigation aimed to test directly if HCN channels may be considered more general targets for NO-cGMP signalling in neurones, bearing in mind that the channels are commonly viewed as being modulated primarily by cAMP ([@b14]; [@b34]; [@b54]).
Addressing this straightforward aim proved frustrating but we can arrive at two main conclusions. The first is that the usual whole-cell recording technique can obliterate the signalling pathway under study. This was not occurring at the level of the HCN channels, which remained active and modifiable by cyclic nucleotides. One problem was the composition of the electrode solution, which was highly inhibitory to NO activation of guanylyl cyclase because of the abnormally high amount of ATP, lack of Ca^2+^ buffering and lack of excess Mg^2+^. As the solution is a conventional one for whole-cell recording from neurones and other cells, it remains possible that its use effectively eliminates a widespread cell--cell signalling pathway, regardless of the precise downstream mechanism. Nevertheless, despite the problems with the electrode solution, it transpired that the over-riding factor was the whole-cell recording technique itself. In this respect, it may be relevant that most previous publications dealing with NO/cGMP regulation of HCN channels in central neurones have used less disruptive sharp electrodes ([@b41]; [@b1]; [@b42]). An exception was in substantia gelatinosa neurones where *I*~h~ was shown to be positively regulated by an NO donor and 8-bromo-cGMP under whole-cell voltage-clamp conditions ([@b30]). Similarly, NO was able to shift the activation curve of HCN channels in the positive direction in sino-atrial nodal cells using a whole-cell solution of similar composition to our 'physiological' solution ([@b10]). The reason why whole-cell recording is disruptive in some cells and not others remains obscure but may be related to the location of the NO transduction unit relative to the recording electrode.
The second conclusion from the recordings from DCN neurones is that HCN channels may be regarded as targets for NO released from either capillary endothelial cells ([@b18]) or nerve fibres in close juxtaposition ([Fig. 2E and F](#fig02){ref-type="fig"}); studies to test these possibilities are in progress. A limitation of the investigation (along with all others performed to date) is that it cannot be excluded that the effect of NO is indirect, resulting from the NO-dependent release of another mediator whose action is transduced through HCN channels. We attempted to address this issue by recording from neurones freshly isolated from the DCN using the method of [@b50] but, in all cells tested (*n* = 6), no *I*~h~ was visible, although voltage-dependent Na^+^ and K^+^ currents were clearly seen (not illustrated). Presumably, the channels are located on the more distal processes that are shorn off during the isolation procedure, an interpretation consistent with immunocytochemical findings that the HCN channel protein in the DCN is concentrated in the neuropil ([@b40]). Nevertheless, studies on isolated heart pacemaker cells have shown that NO (through cGMP) can directly affect HCN channels ([@b25]). The lack of HCN channels in the isolated neurones was also disadvantageous from another perspective, in that it would have allowed the delivery of known NO concentrations directly to the cells under study to discover the levels required to engage this pathway. The studies using intact brain slices give no direct information on this point because NO is inactivated by the tissue extremely rapidly (predicted half-life of \< 10 ms at 10 n[m]{.smallcaps} NO and below), which means that very steep NO gradients will exist across the tissue and therefore that the concentration experienced by a given neurone will depend on its position ([@b24]). Hence, this remains an outstanding issue, and not only for neuronal channels. All that may be stated is that, based on the relative electrophysiological effects of PAPA/NO and DEA/NO in hippocampal slices, the NO concentrations activating *I*~h~ in the perforated-patch recordings using PAPA/NO (10--100 μ[m]{.smallcaps}) as the donor are similar to those required to substitute for endogenous NO in synaptic plasticity ([@b11]; [@b26]) and so are likely to approximate to physiological NO concentrations.
The molecular make-up of the HCN channels operating in the DCN neurones remains unclear. At the mRNA level, HCN2 and HCN4 are highly expressed in the nuclei compared with HCN1 ([@b34]; [@b39]; [@b49]; [@b13]), whereas immunohistochemistry suggests the presence of protein for all three subunits ([@b40]). The mean *V*~1/2~ values and time constants in our study are similar to native channels in thalamocortical and hippocampal neurones that express different combinations of HCN1, 2 and 4 subunits ([@b49]; [@b51]); the *V*~1/2~ values are intermediate to values reported previously for homomeric HCN2 and HCN4 channels ([@b2]). Potentially, therefore, heteromeric channels underlie the currents in DCN neurones. Although HCN channels are commonly regarded as being regulated by the more potent cyclic nucleotide, cAMP, the responsiveness of DCN neuronal channels to NO may reflect the presence of guanylyl cyclase in the vicinity of the channels, producing cGMP in the low micromolar range needed to be effective ([@b14]; [@b34]; [@b54]; [@b58]). Such a possibility appears feasible as guanylyl cyclases in cells can generate up to 100 μ[m]{.smallcaps} cGMP/s in response to NO ([@b38]). Alternatively, some native channels may be less discriminating, considering evidence from the optic nerve and sensory ganglia that the 8-bromo derivatives of cAMP and cGMP were similarly potent HCN channel agonists when applied extracellularly ([@b27]; [@b18]).
One functional outcome of applying NO to the DCN neurones was a clear reduction in the rebound depolarizations occurring post-hyperpolarization. As TTX-sensitive Na^+^ channels were blocked, these depolarizations are presumably carried by Ca^2+^channels. These could be low-threshold T-type Ca^2+^channels ([@b8]) that become deinactivated as a result of the hyperpolarization. However, it should be noted that the role of T-type channels in rebound spiking has recently been questioned when the depth of hyperopolarization does not extend beyond the chloride reversal potential; it may be that different types of Ca^2+^channels and/or modulatory mechanisms are involved ([@b56]; reviewed in [@b57]). Furthermore, it cannot be excluded that the decrease in the peaks could be another effect of NO. For example, there is some evidence in the literature of the acute inhibition of other (L-type) Ca^2+^channels by cGMP and cGMP-dependent kinase ([@b44]; [@b33]). Information on the effects of cGMP on T-type channels is sparse, however, and relates to transcriptional regulation ([@b55]). A simpler explanation is that the changes in rebound depolarization in the presence of NO are secondary to changes in *I*~h~ because this current contributes importantly to the rebound depolarization and associated activation of T-type currents in these ([@b3]) and other ([@b37]) neurones. By analogy with the effect of noradrenaline on thalamic neurones ([@b36]), when *I*~h~ is enhanced, in this case by NO, the reduced hyperpolarization and increased membrane conductance ([Fig. 4](#fig04){ref-type="fig"}) together serve to inhibit T-type channel activation and, hence, inhibit the rebound spiking ([Fig. 5](#fig05){ref-type="fig"}). In DCN neurones, rebound spiking is particularly prominent following repeated stimulation of the inhibitory (GABAergic) Purkinje cell input and, by modifying the firing pattern of the neurones depending on the type of excitatory input to Purkinje cells, the rebound depolarization has been hypothesized to play a role in motor behaviour ([@b3]). Our results suggest a locus for the modulation of such behaviour by NO. A caveat to this interpretation of *in-vitro* data is that recent findings have questioned the importance of the rebound spiking in DCN neurones *in vivo* ([@b7]).
The degree of rebound depolarization may also have repercussions for synaptic plasticity between Purkinje cell axons and DCN neurones, in that a strong depolarization associated with high-frequency firing of action potentials tends to generate long-term potentiation, whereas a suppression of this phase produces long-term depression ([@b4]). The rebound depolarization also appears to play a central role in governing long-term potentiation of excitatory inputs to the DCN neurones ([@b43]). Combined with our results, this leads to the speculation that the presence of NO would favour long-term depression of Purkinje cell axon synapses and a suppression of long-term potentiation at excitatory synapses. Hence, by modulating HCN channels that help shape the post-hyperpolarization changes, the NO-cGMP pathway may play an important role in modulating synaptic plasticity in the DCN and thereby contribute to certain types of motor learning. For example, it has been shown that decreases in endogenous NO in the interposed nucleus of the DCN delayed classical eyelid conditioning ([@b6]).
The generality of our findings carried out using the DCN remains to be investigated but it is noteworthy that cyclic nucleotide-sensitive HCN channels are widely distributed throughout the central nervous system ([@b40]) and signal transduction through the channels could help explain a number of observations of NO acting to modify neuronal membrane potentials or excitability and, potentially, synaptic plasticity ([@b17]). An interaction with HCN channels may also contribute to the participation of the NO-cGMP pathway in certain clinical conditions, such as pain ([@b32]; [@b28]).
This work was supported by The Wellcome Trust (Prize Studentship no. 072330 and Programme Grant nos 066129 and 081512). We are indebted to Barrie Lancaster and Andrew Batchelor for assistance and advice regarding the electrophysiology, and to Kathryn Harris and Giti Garthwaite with respect to the immunohistochemistry.
DCN
: deep cerebellar nuclei
DEA/NO
: 2-(*N*,*N*-diethylamino)-diazenolate-2-oxide
HCN
: hyperpolarization-activated cyclic nucleotide-modulated cation
*I*~h~
: hyperpolarization-activated current
nNOS
: neuronal nitric oxide synthase
NO
: nitric oxide
PAPA/NO
: (Z)-1-\[*N*-(3-ammoniopropyl)-*N*-(*n*-propyl)-amino\]diazen-1-ium-1,2-diolate
TTX
: tetrodotoxin
[^1]: Re-use of this article is permitted in accordance with the Terms and Conditions set out at <http://www3.interscience.wiley.com/authorresources/onlineopen.html>
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Studies investigating rider posture during horseback riding are beginning to emerge in the literature with a common theme being asymmetry in riders ([@b24-jhk-42-113]; [@b10-jhk-42-113]; [@b38-jhk-42-113]; [@b32-jhk-42-113]; [@b11-jhk-42-113]). Since the ability to train the horse to be ambidextrous is considered highly desirable, rider asymmetry is recognized as a negative trait. In studying rider asymmetry, the challenge is not only to measure right-left differences, but also to determine whether they are due to structural differences in anatomical dimensions, inherent laterality, or whether they have been acquired as a consequence of riding. Acquired asymmetry in riders may develop from functional or dynamical differences when performing habitual tasks during riding, which are exacerbated through repetition and/or pain avoidance, resulting in an accumulation of postural defects.
In riding, the complexities of functional and dynamical asymmetry are potentially increased as both horse and rider can be affected, and this may also be magnified by the ability of the rider. Improved harmony between the horse and the rider has been reported in more experienced riders ([@b27-jhk-42-113]; [@b24-jhk-42-113]; [@b28-jhk-42-113]; [@b8-jhk-42-113]), so it is reasonable to expect that highly experienced riders and their horses would be more symmetrical. There is evidence, however, to suggest that even highly trained dressage riders find it difficult to form an accurate perception of rein tension in left and right hands, due to the amount of sensory information and external stimuli they are receiving simultaneously ([@b39-jhk-42-113]).
Laterality due to dominance of one side of the body is another cause of asymmetry in humans. Substantial inter-limb differences in control of limb dynamics are associated with handedness in adults ([@b5-jhk-42-113]). In sport, differences in performance due to limb dominance are commonly reported in the literature ([@b19-jhk-42-113]). Typically, the preferred limb is used for mobilization whereas the non-preferred limb is used for support ([@b35-jhk-42-113]). Postural adaptations, such as asymmetry of the pelvis are also found ([@b7-jhk-42-113]). It appears that modified tension patterns within the musculo-ligamentofascial apparatus stabilizing the pelvis, due to left-right differences in the mechanical loads transmitted through it, increase the potential of developing static pelvic asymmetry ([@b13-jhk-42-113]). During riding, propulsive forces from the horse's limbs are transmitted to the rider ([@b40-jhk-42-113]) and asymmetry in these forces due to left-right differences in strength or subclinical lameness could affect loading symmetry on the rider's pelvis. The magnitude and prevalence of pelvic asymmetry in unilateral sports was, however, found to be greater than in bilateral sports ([@b7-jhk-42-113]).
Pelvic asymmetry in the frontal plane may also occur due to leg length discrepancies, whereby paired limbs are noticeably disproportionate, although agreement of what constitutes a clinically significant length difference sufficient to induce postural changes is currently lacking ([@b6-jhk-42-113]; [@b12-jhk-42-113]; [@b15-jhk-42-113]). When these changes are induced temporarily using a lift in the shoe, in addition to inducing pelvic tilt, they are also reported to cause pelvic torsion and increased lateral flexion of the trunk towards the side of the limb that has been lifted ([@b42-jhk-42-113]). [@b38-jhk-42-113] found a relationship between asymmetry in shoulder rotation during riding and a leg length discrepancy during standing. They used the palpation meter (PALM) method ([@b29-jhk-42-113]) to measure leg length indirectly, so it is possible that pelvic asymmetry rather than a leg length inequality influenced shoulder rotation.
Subtle anatomic abnormalities in the pelvis are also associated with altered mechanics in the lumbar spine, often due to lower back pain ([@b1-jhk-42-113]). Back pain has been linked to sagittal plane spinal posture ([@b26-jhk-42-113]), which was suggested to be manifested as a loss of lordosis and an anterior shift in the sagittal vertical axis of the body. These postural changes were related to the development of degenerative changes in the spine ([@b33-jhk-42-113]). It is unclear though whether individuals develop altered static and dynamic loading patterns prior to or subsequent to the first bout of pain ([@b17-jhk-42-113]). Subjects with lower back pain are also reported to have deficits in standing and seated balance and automatic postural coordination ([@b9-jhk-42-113]). In riders the incidence of lower back pain was reported to be higher than the incidence found in the general population ([@b21-jhk-42-113]; [@b22-jhk-42-113]), but it is unclear whether risks are discipline specific ([@b30-jhk-42-113]; [@b22-jhk-42-113]).
Symmetry in riders is desired, but in a small group of riders anatomical and functional asymmetry has previously been identified ([@b38-jhk-42-113]). Testing a larger population of riders to establish if particular asymmetry is commonly found, if anatomical or/and functional asymmetry is more prevalent in less experienced riders compared to higher level riders ([@b27-jhk-42-113]; [@b24-jhk-42-113]; [@b28-jhk-42-113]; [@b8-jhk-42-113]) or riders who have been riding for a greater number of years ([@b30-jhk-42-113]), would make an important contribution to current knowledge. The aim of this study was therefore to determine whether anatomical asymmetry (leg length, pelvis and shoulder height), functional asymmetry (trunk lateral bending and axial rotation range of motion (ROM) during sitting) and dynamical asymmetry (grip strength) were prevalent in a larger population of riders and to determine whether typical traits exist due to riding. It was hypothesised that a greater number of years riding in riders with less ability would lead to an accumulation of different types of asymmetry, together with increased prevalence of postural defects and pain. If this is indeed the case, further studies would be warranted to develop educational strategies to decrease these risks.
Material and Methods
====================
Ethical approval was obtained for this study from University of Central Lancashire under the approval number PSY1011104. Written informed consent was obtained from all riders prior to commencement of the study.
Participants
------------
The study group was comprised of 132 females and 2 males that attended a British Dressage Camp in the UK in 2011 or that attended a test event at the Michigan State University in the USA in 2011. Information was provided to participants prior to attending each testing event and participants were recruited on a voluntary basis. Of the group, 127 riders were right handed, 5 were left handed and 2 were ambidextrous. As the group was not typically representative of the normal population in relation to handedness ([@b3-jhk-42-113]), only data from the right handed participants was retained. This also allowed inherent right-handed traits to be considered in the analysis of results. The statistics for the group are shown in [Table 1](#t1-jhk-42-113){ref-type="table"}. The background of each participant was obtained with the use of a questionnaire to determine their injury history, prevalence of pain, their lifetime involvement with riding and other sports and their hand dominance. A competition level was categorized based on the level ridden in competition according to the British Dressage or the United States Dressage Federation tests: 1) intro/prelim or training/first; 2) novice or second, 3) elementary or third level and above. Riding in years was categorized as: 1) 1--19, 2) 20--39, 3) 40+. Body height was measured in standing posture against a wall and body mass was measured with a weighing scale (Salter, UK). Leg length was measured in a prone position with a neutral pelvis. Measurements were taken three times on each leg in a random order from the anterior superior iliac spine (ASIS) to the distal edge of the medial malleolus using a flexible tape measure ([@b2-jhk-42-113]). This method has acceptable validity and reliability as a screening tool for assessing leg length discrepancy provided an average of two tape measurements is used ([@b15-jhk-42-113]).
Procedures
----------
Images from testing are shown in [Figure 1](#f1-jhk-42-113){ref-type="fig"}. Sagittal plane images of standing posture were captured using the method described by [@b26-jhk-42-113]. The overall postural type was categorized ([@b26-jhk-42-113]; [@b37-jhk-42-113]) by a chartered physiotherapist. Grip strength was measured using a grip strength dynamometer (Takei, Japan). Participants were asked to grip the dynamometer as hard as they could whilst flexing their elbow at 90 degrees ([@b16-jhk-42-113]). Three trials were performed in a random order between left and right arms.
The UK studies used a four camera infrared motion capture system for which the error in a linear measurement of 750.5 mm was \<2.3mm (Qualisys Capture Systems, Gothenburg, Sweden) and the US studies used a ten camera infra-red motion capture system for which the error in a linear measurement of 1000 mm was \<0.8 mm (Motion Analysis Corp., Santa Rosa, CA). Participants were recorded during seated posture and whilst performing trunk motion exercises. The systems were calibrated in order that a horse model with a saddle could be placed along one of the horizontal axes, allowing absolute positions relative to the laboratory coordinates to be measured. Of the participant group, 94 riders (1 male, 93 female), age 38.7 ± 10.8 years participated in standing and sitting postural measurements and trunk ROM tests. Retro-reflective markers were attached to the left and right acromion process, iliac crest, posterior superior iliac spine (PSIS), greater trochanter and a cluster of four markers was firmly attached in the upper thoracic region. A standing trial was captured initially with participants in the anatomical position in order for the tracking markers to be referenced to the anatomical markers. All markers remained in place for the duration of the testing.
Markers attached to the participants were then captured during sitting in their normal riding posture on a dressage saddle that was secured to a horse model. To measure trunk flexibility a lightweight wooden pole was placed across the shoulders to prevent excessive motion of the shoulder girdle. Participants remained seated on the saddle horse and performed slow left and right lateral bending and left and right rotation movements to the end of their ROM returning each time to a neutral position. Three trials for each movement were captured at 100 Hz in a random order using the same slow movement pattern for each movement (as instructed by the investigator), based on the procedures used by [@b1-jhk-42-113]. Markers tracking the trunk were identified in the software Qualisys Track Manager (Qualisys Capture Systems, Gothenburg, Sweden) and Cortex (Motion Analysis Corp., Santa Rosa, CA) then exported into Visual 3D (C-Motion Inc., Germantown, MD).
A kinematic model was created for the trunk and pelvis and applied to the sitting posture and trunk flexibility data. The acromion process and iliac crest markers were used to define the proximal and distal ends of the trunk and the cluster of four markers on the upper thorax were used to track trunk movements. The iliac crest and greater trochanter markers defined the proximal and distal ends of the pelvis and these markers together with the PSIS markers were used to track the pelvis. A 4^th^ order Butterworth filter ([@b31-jhk-42-113]) with cut off frequency of 5 Hz was applied to markers tracking trunk motion during the trunk flexibility tests to remove higher frequency noise within the data. A 5 Hz cut off was chosen as this retained 95% of the signal power. For standing and seated posture absolute position of the markers on the left and right acromion processes and iliac crests in the vertical direction were extracted. For trunk lateral bending and rotation, the range of motion of the trunk relative to the pelvis was extracted using an XYZ Cardan sequence, where X was flexion-extension, Y was lateral bending and Z was rotation. The sign convention was based on the right handed rule: in lateral bending right shoulder downwards was positive (right lateral bending) and in rotation right shoulder rotating anticlockwise when viewed from above was positive (right rotation).
Statistical Analysis
--------------------
Two factors were investigated: number of years riding (three levels) and competition experience (three levels). For the measured variables (leg length, grip strength, height of the acromion processes and iliac crests during standing and seated posture, lateral bending ROM and rotation ROM) the magnitudes for right (+) and left (-) sides were determined and the absolute difference between right and left sides was calculated (right -- left). Data was tested for normality using Kolmogorov-Smirnov normality tests. All variables were normally distributed, except for leg length, which was transformed. To study the prevalence of acquired asymmetry in riders a 2 × 3 ANOVA was used to determine significant differences between two factors: years riding (× 3 levels) and competition experience (× 3 levels) with age included in all tests as a covariate except for leg length. Leg length was considered to be a skeletal difference and therefore was not expected to have been acquired due to age. Posture classification was tabulated in accordance with pain reported in specific parts of the body. In addition, the prevalence of pain in relation to posture (% per group) and the prevalence of postural types that were not normal (% per group) were calculated.
Results
=======
Statistics for the whole group are shown in [Table 1](#t1-jhk-42-113){ref-type="table"}. For this rider group 51% rode pure dressage only, 49% rode in dressage and other equestrian sports, 71% were currently participating or had previously participated in other non-equestrian sports or types of exercise, 55% reported having had at least one serious injury prior to commencing the study.
Significant anatomical asymmetry was found for the difference in standing acromion process height for the competition level (−0.07±1.50 cm Intro/Prelim; 0.02±1.31 cm Novice; 0.43±1.27 cm Elementary+; p=0.048) and for sitting iliac crest height for years riding (−0.23±1.36 cm Intro/Prelim; 0.01±1.50 cm Novice; 0.86±0.41 cm Elementary+; p=0.021). For functional asymmetry, a significant interaction was found for lateral bending ROM for years riding × competition level (−1.17±6.78 deg Intro/Prelim; 2.64±6.27 deg Novice; 2.37±5.05 deg Elementary+; 0.84±5.32 deg 0 to 19 yrs; 0.23±6.99 deg 20 to 39 yrs; 3.13±5.66 deg 40+years; p=0.047). No other significant differences were found between right and left sides, although grip strength was notably higher on the right for all groups. For all variables the coefficient of variability was high, see [Tables 2](#t2-jhk-42-113){ref-type="table"}, [3](#t3-jhk-42-113){ref-type="table"} and [4](#t4-jhk-42-113){ref-type="table"}. The posture classifications of the riders according to regions of pain they reported are shown in [Table 5](#t5-jhk-42-113){ref-type="table"}. The prevalence of pain for riders with postural defects, which constitutes all other postures except for normal posture, and also for riders with normal posture, shown by group is reported in [Table 6](#t6-jhk-42-113){ref-type="table"}. A trend is seen in riders with postural defects developing back and/or neck pain with an increasing level of competition.
Discussion
==========
This study assessed a large sample of right-handed riders in relation to anatomical, functional and dynamical asymmetry to determine whether typical traits existed due to riding. The interaction in functional asymmetry with lateral bending ROM to the right greater for years riding and the competition level only in part supports the hypothesis, as symmetry was expected to improve with the competition level. For significant anatomical asymmetry, the mean difference in standing acromion process height increased with the competition level, which did not support the hypothesis. The mean difference in sitting iliac crest height altered from higher on the left to higher on the right with years riding, so there was evidence that sitting pelvic asymmetry may develop in riders, but this was not influenced by ability. A trend of increased prevalence of pain in riders at higher competition levels was found in riders that had postural defects, but this did not consistently increase with years riding.
The interaction in functional asymmetry for lateral bending ROM and an increase in prevalence of pain from 38% in low level riders to 59% in high level riders with postural defects may be clinically important. Symmetry in lateral bending and rotation was suggested to be clinically important within the individual, particularly for the diagnosis of lower back pain ([@b1-jhk-42-113]). Higher lumbar motion asymmetry has previously been reported in lower back pain patients compared to control subjects, although asymmetry is still evident in the normal population ([@b14-jhk-42-113]; [@b1-jhk-42-113]). Pain has been reported in higher level riders ([@b22-jhk-42-113]), which has been attributed to the requirement for them to absorb the considerable vertical movement of the centre of mass of the horse whilst sitting in an upright dressage posture ([@b4-jhk-42-113]). However, [@b22-jhk-42-113] found no conclusive evidence of pathologies in riders with lower back pain. In the context of pain, lateral bending asymmetry may be a restriction or stiffness in either vertebral or paravertebral structures ([@b1-jhk-42-113]), so it is possible that pain in riders is largely related to sub-clinical asymmetry. Higher level riders with postural defects may therefore have more difficulty absorbing the movements of the horse, resulting in greater pain and increased muscle stiffness. Conversely, pain avoidance during riding may increase the prevalence of postural defects and muscle imbalances in higher level riders. Further work is needed to investigate the cause and effect relationship between back pain and horse riding.
A greater mean standing acromion process height on the right for the most experienced rider group may also be linked to differences in lateral bending ROM. [@b36-jhk-42-113] suggested that greater muscle development and therefore muscle stiffness on the right side would limit lateral bending to the left. Certainly, right grip strength was greater for all groups, which would be expected for a right handed population ([@b41-jhk-42-113]) and grip strength was correlated with muscle mass ([@b18-jhk-42-113]). [@b25-jhk-42-113] speculated that the use of the dominant hand in daily activities may train muscle fibres towards the properties of fast-twitch fibres and more efficient control of intersegmental dynamics may also alter muscle development ([@b5-jhk-42-113]). Muscle hypertrophy on the right could therefore explain the increase in acromion process height, which may also explain the reduction in left lateral bending ROM. This finding is not typical, as Kendall et al. (1983) reported that the dominant shoulder was normally positioned lower than the non-dominant shoulder in most people. The dynamic control needed to provide suitable signals to the horse whilst maintaining upright upper body posture is extremely important when riding dressage and it is known that rein tension varies between right and left hands ([@b23-jhk-42-113]). Different left-right muscle recruitment patterns may be used to produce similar signals to the horse, which result in asymmetrical muscle development, asymmetrical shoulder height and asymmetry in lateral bending ROM.
Another explanation for reduced lateral bending ROM to the left would be a restriction due to a left axial rotation postural position of the trunk whilst completing the exercise ([@b36-jhk-42-113]). Greater left axial rotation ROM was evident for 40 + years riding and the Elementary + level, although these were not significant and no interaction was found. Also, a difference in axial rotation ROM may not relate to a postural rotational position during a lateral bending exercise. [@b38-jhk-42-113] found a preferred posture of right rotation of the shoulders during riding not left rotation, so again, this does not support the idea that riders may commonly have a left axial rotation postural defect. More quantitative measurements of anatomical, functional and dynamical asymmetry compared to ridden postural asymmetries are needed to understand the effects of riding on strength, posture and flexibility.
Finally, [@b1-jhk-42-113] suggested that lateral bending ROM was highly associated with pelvic asymmetry in the normal population. As iliac crest height during sitting was only influenced by years riding and not ability it is unlikely that pelvic asymmetry influenced lateral bending ROM. [@b7-jhk-42-113] suggested that pelvic asymmetry due to lateral dominance may decrease when athletes undertake bilaterally dominant activities. Evidence from this study suggests that sitting pelvis height does alter over time, but this does not necessarily constitute an improvement in symmetry. Riders are often reported to collapse their hip to one side and may show increased pressure under the saddle on the same side or on the opposite side (Clayton, unpublished). It may be surmised that riding posture and the application of signals to the horse influence static sitting posture. [@b13-jhk-42-113] found changes in functional asymmetry of the lumbo-pelvo-hip complex following mechanical stimulation. This was attributed to a change of tension patterns within the musculo-ligamento-fascial apparatus that maintains stability of the pelvis. It is possible that together with the application of signals to the horse, the force transmission through the pelvis during riding alters the tension in the musculo-ligamento-fascial apparatus over time and, therefore, influences pelvic alignment.
Although an anatomical difference and not an acquired difference, a leg length discrepancy was not found in this rider sample, which is contrary to the findings of [@b38-jhk-42-113]. They reported a shorter right leg in a considerably smaller sample of riders. Discrepancies can occur through an inaccurate measurement of the limbs, depending on the method used ([@b34-jhk-42-113]). Our leg length findings are also supported by standing iliac crest height results, as iliac crest height was reported to provide clinically useful evidence for suspecting leg length inequality ([@b42-jhk-42-113]).
The absolute difference between left and right sides for each measurement and each participant was calculated, grouped, and then used to establish asymmetry patterns in riders. Some participants tended to produce larger values to the right and other participants tended to produce larger values to the left, so the mean for each group tended towards zero and the standard deviation tended to be large. This is highlighted by the large values reported for the coefficient of variability in [Tables 2](#t2-jhk-42-113){ref-type="table"}, [3](#t3-jhk-42-113){ref-type="table"} and [4](#t4-jhk-42-113){ref-type="table"}. Due to the large variability between riders further work should incorporate longitudinal within-rider monitoring, which may indicate better the cause and effect relationships between riding and changes in symmetry.
This study reported anatomical, functional and dynamical asymmetry in a large sample of riders. Data were collected at a number of locations in the UK and USA and as such care was taken to replicate the same procedures at each data collection session. Despite this there are differences in the accuracy of the measurements taken using motion capture techniques, as identified in the method. To reduce bias between samples and compare symmetry the difference between right and left measures for each variable were analysed. For the motion capture data markers were missing on some occasions during testing, mainly due to occlusion, so for some participants the end of range of motion could not be determined. In addition, on a small number of occasions participant information or recorded measurements were missing. Consequently, the number of samples included for each variable was not consistent throughout the study. These are reported in [Tables 2](#t2-jhk-42-113){ref-type="table"}, [3](#t3-jhk-42-113){ref-type="table"} and [4](#t4-jhk-42-113){ref-type="table"}.
Conclusions
===========
Symmetry of posture, strength and flexibility was assessed in a large population of riders to determine whether typical traits existed due to riding. Lateral bending ROM to the left was reduced in higher level riders that had ridden for a longer amount of time. This may be attributed to asymmetric shoulder height, suggesting that strength and therefore muscle development is greater on the right side of the body. Alternatively there is evidence to suggest that this may relate to pain. A difference in sitting pelvic asymmetry was found for years riding, which may also have restricted lateral bending ROM in the higher level riders. The demands on dressage riders competing at higher levels may predispose these riders to a higher risk of developing asymmetry and potentially chronic back pain rather than improving their symmetry.
The authors would like to acknowledge British Dressage Regional Development Officers (Katharine Perry, Lou Jones, Sarah Huntridge and Amanda Heath) and Jane Walker (Active Rider) for their assistance in arranging data collection days at the 2011 summer camps for this project. The authors would also like to thank LeeAnn Kaiser, Lila Zarski, Narelle Stubbs and Nicole Rombach (Michigan State University) for assistance with data collection and processing in the USA.
{#f1-jhk-42-113}
######
Descriptive statistics of the riders
Mean
------------------------ --------------
Age (years) 39.2 (12.1)
Number of years riding 26.3 (12.3)
Body height (cm) 166.0 (12.8)
Body mass (kg) 67.1 (12.8)
BMI 24.4 (4.1)
Reach distance (cm) 16.5 (8.8)
######
Statistics for anatomical asymmetry: Mean for the left and right sides, difference (Right-Left), coefficient of variability (%COV), significance (P) and number of participants included in the group (n) and including left and right measurements for leg length, acromion process (AP) height (standing and sitting) and iliac crest height (standing and sitting).
^\*^Age not used as a covariate for leg length
**Leg length^\*^** L (cm) Difference (R-L) cm R (cm) \% COV p n
-------------------- ------------ ------------------------ ------------------------ ------------ ------------- ----------- ----
Yrs Riding 0--19 83.4 (6.7)  83.4 (6.8) 2050 0.676 27
20--39 83.7 (5.8)  83.7 (5.9) 1450 54
40+ 84.7 (5.1)  84.8 (5.2) 643 38
Comp Level Int/pre 82.1 (6.1)  81.8 (6.1) 264 0.349 34
Nov 83.6 (5.3)  83.7 (5.4) 6300 44
El+ 85.8 (5.5)  85.5 (5.6) 7700 45
**Standing AP**  Interaction 0.924
Yrs Riding 0--19 144 (12.7)  145 (12.7) 444 0.480 20
20--39 146 (15.1)  146 (15.1) 800 39
40+ 151 (13.9)  151 (13.8) 923 36
Comp Level Int/pre 146 (12.4)  145 (12.4) 2143 **0.048** 30
Nov 147 (14.5)  147 (14.7) 650 35
El+ 153 (18.6)  153 (18.3) 302 33
**Sitting AP**  Interaction 0.792
Yrs Riding 0--19 146 (8.6)  146 (8.5) 263 0.712 19
20--39 146 (10.9)  146 (11.0) 2600 40
40+ 148 (11.4)  148 (11.3) 1150 36
Comp Level Int/pre 146 (8.8)  145 (8.6) 1029 0.513 30
Nov 147 (10.8)  147 (10.6) 2317 35
El+ 151 (17.6)  151 (18.0) 811 33
**Standing iliac**  Interaction 0.891
Yrs Riding 0--19 108 (10.6)  107 (10.8) 327 0.070 20
20--39 109 (13.0)  108 (13.1) 209 37
40+ 113 (12.2)  113 (12.2) 790 35
Comp Level Int/pre 109 (10.2)  108 (10.5) 225 0.489 30
Nov 110 (12.9)  110 (13.2) 5533 33
El+ 111 (14.1)  115 (17.6) 1586 32
**Sitting iliac**  Interaction 0.212
Yrs Riding 0--19 111 (6.2)  111 (6.0) 296 **0.021** 15
20--39 112 (9.9)  112 (9.7) 457 32
40+ 112 (10.7)  112 (9.5) 200 30
Comp Level Int/pre 112 (6.5)  111 (6.6) 705 0.919 24
Nov 111 (9.1)  111 (9.0) 15300 28
El+ 117 (19.2)  117 (19.0) 1455 28
 Interaction 0.925
######
Statistics for dynamical asymmetry: Difference (Right-Left), coefficient of variability (%COV), significance (P) and number of participants included in the group (n) and including left and right measurements for grip strength
**Grip strength** L (kg) Difference (R-L) kg R (kg) \% COV p n
------------------- ------------- ------------------------ ------------------------ ------------- ------- ------- ----
Yrs Riding 0--19 23.6 (7.9)  25.2 (7.8) 216 0.427 27
20--39 24.6 (8.7)  26.5 (9.1) 164 54
40+ 23.2 (10.0)  24.9 (10.7) 195 38
Comp Level Int/pre 25.1 (8.6)  26.0 (8.9) 302 0.064 34
nov 23.5 (7.4)  25.2 (7.5) 180 44
el+ 23.1 (10.2)  25.7 (11.0) 135 45
 Interaction 0.653
######
Statistics for functional asymmetry: Difference (Right-Left), coefficient of variability (%COV), significance (P) and number of participants included in the group (n) and including left and right measurements for lateral bending and axial rotation range of motion
**Lateral bending** L (deg) Difference (R-L) deg R (deg) \% COV p n
--------------------- ------------ ------------------------ ------------------------ ------------- ----------- ------- ----
Yrs Riding 0--19 38.9 (8.3)  39.8 (7.3) 631 0.181 21
20--39 37.2 (8.4)  37.6 (8.9) 1795 38
40+ 34.2 (7.7)  37.6 (7.0) 182 34
Comp Level Int/pre 38.5 (8.4)  37.6 (7.8) 817 0.143 29
nov 36.5 (8.6)  39.1 (7.2) 238 35
el+ 35.0 (7.6)  37.3 (8.5) 215 32
**Rotation**  Interaction **0.047**
Yrs Riding 0--19 45.2 (8.6)  43.0 (7.5) 260 0.528 17
20--39 40.9 (8.7)  42.0 (9.1) 842 34
40+ 43.9 (7.0)  41.7 (7.5) 347 34
Comp Level Int/pre 43.6 (7.1)  41.9 (8.3) 493 0.963 24
nov 43.1 (9.0)  43.1 (7.5) 87000 33
el+ 43.5 (8.9)  41.6 (8.1) 415 31
 Interaction 0.813
######
Riders categorised by pain and posture type (n=122). Postural data was omitted from the remaining 5 riders as no information relating to pain was provided
**Posture** **No Pain** **Lumbar** **Thoracic** **Shoulder** **Neck** **Various** **% of riders**
------------------- ------------- ------------ -------------- -------------- ---------- ------------- -----------------
Lordotic 10 3 0 1 1 6 17
Kyphotic/lordotic 12 8 2 5 3 8 31
Kyphotic 2 2 1 0 1 2 7
Swayback 3 5 0 2 1 4 12
Normal 9 9 0 5 2 9 28
Flatback 3 1 0 1 0 1 5
**Total** **39** **28** **3** **14** **8** **30** **100**
######
Prevalence of pain for riders with postural defects (all other postures except for normal posture) and for riders with normal posture presented as a % of the group and prevalence of postural defects (all other postures except for normal posture) presented as a % of the group (n=122)
**Level**
------------------------------------ ------------ ----------- -------- -------- --------
**Postural Defects** **0--19** 23 55 33 **29**
**20--39** 60 38 74 **56**
**40+** 17 54 47 **45**
**Total** **38** **47** **59**
**Normal Posture** **0--19** 15 9 33 **11**
**20--39** 7 24 11 **15**
**40+** 50 31 26 **32**
**Total** **18** **22** **20**
**Prevalence of postural defects**
**0--19** 54 91 67 **70**
**20--39** 87 71 84 **80**
**40+** 50 69 68 **63**
**Total** **68** **76** **73**
[^1]: Authors submitted their contribution of the article to the editorial board.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#S1}
============
Healthy People 2020, the federal public health agenda, has set a goal of "Increase(ing) the proportion of persons with hemoglobinopathies who receive disease-modifying therapies"^[@R1]^. For the vast majority of people with sickle cell disease (SCD), the Healthy People goal will be reached through increased use of hydroxyurea (HU). Critical questions surrounding its use include how this drug works to ameliorate the clinical severity of SCD and what sub-population of children with SCD benefit most from its use. This review addresses these questions from a translational science perspective.
Sickle cell disease (SCD) affects an estimated 90,000 people in the U.S.^[@R2]^ with over 1900 newborns detected annually through universal newborn screening^[@R2]^. Infant screening, early preventive therapy and parental guidance have largely eliminated early child mortality from SCD^[@R3]--[@R5]^. Moreover, specialized care and on-going preventive services have prolonged average life expectancy^[@R6]^. Despite these successes, multi-organ damage and mortality accumulate by early adulthood, resulting in shortened lifespan^[@R6]^.
HU holds expanding promise for improved clinical outcomes. Over two decades ago, the seminal Multicenter Study of Hydroxyurea (MSH) phase III trial for adults demonstrated the striking clinical impact of HU: 40% reduction in the incidence of acute pain episodes, acute chest syndrome and hospitalization^[@R7]^. These results led to approval in 1998 of HU for use in symptomatic SCD by the United States Food Drug Administration (FDA). HU remains the only FDA-approved drug for SCD, but approval does not extend to pediatric use. The approval gap for children is partially attributed to the lack of a commercial pharmaceutical sponsor. Helping to span gap is the FDA's recent commissioning of a pediatric study of the pharmacokinetics of HU and its relative bioavailability of the liquid formulation (<http://clinicaltrials.gov/show/NCT01506544>).
Clinical efficacy of HU treatment varies between individuals, although most patients with severe phenotypes benefit from its use^[@R7],[@R8]^. This review describes newly identified mechanisms for the effects of HU, including genetic regulation of fetal hemoglobin (HbF) as a disease modifier and the biologic effects of HU on blood vessels and gene regulation. These recent advances improve the prospects for prospectively assessing efficacy of HU therapy, are inspiring clinical trials for additional salutatory effects of HU and may guide future drug development.
CLINICAL EFFECTS {#S2}
================
The profound clinical effects of HU for children with SCD have been recently reviewed^[@R9]--[@R11]^ summarized here and in [Table 1](#T1){ref-type="table"}. Much of the work on HU in children with SCD has come from phase III trials led by Ware and colleagues, including pivotal studies such as HUGS, HUG-KIDS^[@R12]--[@R14]^, HUSOFT^[@R15]^, BABY-HUGS^[@R16]--[@R18]^ including an early pediatric trial published in 1999^[@R12]^. French investigators have also contributed insights into the impact of HU^[@R19],[@R20]^. Randomized pediatric trials with HU have demonstrated decreased pain episodes^[@R18]^, acute chest syndrome, hospitalization^[@R8],[@R11],[@R18]^, transfusion and splenic auto-infarction^[@R18]^ and improved quality of life^[@R21],[@R22]^. Prolonged use sustains the laboratory effects of decreased anemia, markers of hemolysis, white blood cell and platelet counts and increased red cell mean corpuscular volume (MCV)^[@R23]^. Early HU use stabilizes renal hyperfiltration^[@R24]^ and hyposthenuria^[@R25]^, as well as age-dependent decreased HbF^[@R18]^. Induction of HbF is described below.
Of note, while the laboratory effects of HU apply across the pediatric ages tested, many of the various clinical improvements noted for one age range have not necessarily been assessed for other ranges. For example, reduced dactylitis, hyposthenuria and transfusions were noted in the BABY HUG trial of children enrolled at age 9--13 months^[@R17],[@R18]^. Improved transcranial Doppler blood flow through large cerebral arteries has been demonstrated in school-aged children^[@R26],[@R27]^. Despite positive findings, some of these trials had mixed results. For example, the primary endpoints of the BABY-HUG study were not met^[@R18]^. In the SWiTCH study for secondary stroke prevention, continued chronic transfusion was advantageous over HU with phlebotomy^[@R28]^. Moreover, HU reduces but does not eliminate the symptoms and morbidity of SCD. For example, the SWiTCH trial demonstrated that chronic transfusions more effectively prevented pain episodes than HU with phlebotomy^[@R29]^.
Two long-term studies demonstrated substantially improved lifespans from prolonged use of HU in adults, including a study based on the MSH trial^[@R30],[@R31]^. Prospective lifespan data for children taking HU are not yet available due to later uptake into pediatric trials. Nonetheless, a recent retrospective study from Brazil reported improved childhood mortality for those taking HU for up to six years^[@R32]^. Collectively, these data are increasingly persuasive about the enduring impact of HU on SCD.
The pharmacokinetics of HU appears to follow a bi-phenotypic metabolism in children^[@R33]^. Multiple single base polymorphisms (SNP) are associated with two apparent pharmaco-kinetic profiles of HU uptake and excretion. However, these genotypes do not correlate with response by the biomarker HbF.
FETAL HEMOGLOBIN {#S3}
================
The clinical severity of SCD is highly variable. Children experience multiple different clinical complications of differing severity and frequency. Fetal hemoglobin (HbF) is of critical importance in the major sickle sub-type of HbSS (and HbS-Beta zero thalassemia, here collectively referred to as HbSS). Lower HbF levels correlate with overall more severe disease manifestations^[@R34]^. Unlike HbA, HbF actively inhibits the polymerization of sickle hemoglobin, the underlying patho-physiology of SCD. In solution, fetal hemoglobin concentration higher than 15% prevents sickle globin polymerization^[@R35]^. The cut-off for defining lower risk of severe complications has been estimated at 20%^[@R36]^.
Sharp declines of HbF during infancy occur as HbF-producing gamma globin is replaced by beta globin. This switch leads to the predominant expression of either HbA or HbS. The F-to-S switch in children affected by HbSS^[@R37]^ occurs more gradually than the F-to-A switch in non-anemic children. HbF levels in toddlers with SCD stabilize by age 3 or 4 and are generally constant throughout childhood. Despite bearing the same beta globin sickle variant, affected populations with African ancestry exhibit wide variations in HbF levels^[@R37]--[@R40]^. In the U.S., pediatric levels vary from 3--20% of total hemoglobin, compared to only 0.5--2% for non-anemic individuals. The average HbF level in the U.S SCD pediatric population is approximately 10%^[@R36]^.
HU USE {#S4}
======
Mechanism of action {#S5}
-------------------
The physiology of the HU effect is complex, and can generally be generally categorized into two overlapping pathways: effects on HbF production and improved blood flow through reduced intercellular adhesion ([Figure 1](#F1){ref-type="fig"}). HU is a short-acting cytotoxic drug that induces a state of "stress erythropoiesis." Enhanced HbF production from intermittent mild marrow toxicity is believed to stem from steady shifting of marrow physiology to the stressed state. The marrow responds to the repetitive pharmacologic injury of daily use by enhanced erythropoiesis and increased HbF production^[@R34],[@R41]^. Paradoxically, the net effect of marrow toxicity is induced HbF and stabilization of cellular hemoglobin solubility. These effects lead to decreased RBC membrane damage and hemolysis^[@R34],[@R41]^.
HbF induction usually occurs within the first few months after initiating HU, and is reversible upon cessation or diminution of dosing ([Figure 2](#F2){ref-type="fig"}). Relevance of HU induction of HbF was demonstrated through a proof-of-principle murine model for SCD. Lack of expression of human HbF precluded HU induction in those mice. In that murine model, HU itself had no effect on improving anemia or protecting organs from SCD damage. In contrast, HbF gene therapy markedly improved the blood smear, microscopic and organ-level pathologic effects of SCD^[@R42]^.
HU appears to influence red blood cell-endothelial interactions. Decreased expression of RBC, WBC and endothelial integrins and other adhesion molecules probably improves micro-vascular blood flow and reduces pro-inflammatory cell-cell interactions^[@R43],[@R44]^. Microvascular effects of SCD and HU appeared to be replicated using an interesting microfluidic model of blood flow and endothelialized microfluidic channels^[@R44]^. Whole blood samples from SCD lead to microvascular occlusion and thrombosis. Blood samples from patients with SCD had diminished velocity and greater tendency to obstruct in the microchannels. These effects nearly normalized using blood from patients on HU^[@R44]^. HU may be associated with reduced generation of microparticles, suggesting a reduction in markers of inflammation and thrombosis^[@R45]^.
HU may reduce cellular adhesion in general and/or adhesion provoked by infection or inflammation. Integrins and other cell surface glycoproteins regulate neutrophil migration and red blood cell flow through endothelial interactions. In a murine model for SCD and pneumococcal pneumonia and sepsis, HU provided some protection by decreasing the recruitment of neutrophils into infected lungs. Mice genetically engineered to lack E-selectin were not protected by HU^[@R46]^. This finding strengthens the view that HbF-independent effects of HU include decreasing leukocyte-endothelial adhesion.
HU may also stimulate nitric oxide production as an NO donor or through stimulating intermediates (discussed below). As a potent vasodilator, NO repletion contributes to improved vascular health in SCD ([Figure 1](#F1){ref-type="fig"})^[@R47]^. Along with decreased "sticky" interaction between of blood cells and the endothelium, enhanced nitric oxide-induced local vasodilation may also benefit blood flow ([Figure 1](#F1){ref-type="fig"})^[@R48],[@R49]^. However, questions have arisen regarding these effects from NO^[@R50]^. In all, decreased pathology from damaged red blood cells and pathologic interactions between red blood cells and endothelial cells appear to synergistically reduce clinical signs symptoms and morbidities of the disease ([Table 1](#T1){ref-type="table"}). The ameliorative effects of HU appear to persist for as long as it is taken and the pharmaco-kinetics are maintained.
HbF response to HU {#S6}
------------------
The individual extent of HU-induced HbF is highly variable. Standard pediatric dosing of HU adjusts for dose-dependent myelotoxicity^[@R14],[@R33]^. Under these conditions, HU generally induces HbF an additional 8--18% over baseline levels^[@R14],[@R33],[@R51],[@R52]^. In contrast to the bio-marker HbA1c for diabetes, no absolute HbF target exists. Nonetheless, peak attained HbF levels remain fairly constant in childhood ^[@R23]^. No absolute limit to the therapeutic amount of HbF induction has been described. For example, Southeast Asians or Saudis with SCD have baseline HbF levels averaging 16--20%. HU induction raises their levels 1.5 to 2-fold, associated with further diminution of their already tempered clinical symptoms^[@R53]^.
Children with SCD generally have higher baseline HbF levels than adults and more pronounced HbF response to HU^[@R14],[@R54]^. Factors responsible for differences may include the need for highly regenerative marrow red cell precursors and, for HU, normal renal function for prompt excretion. Adults normally experience age-dependent decreased marrow cellularity. In SCD, disease-related marrow infarcts and other age-related physiologic effects could exacerbate normal marrow regression. Age-related diminution of response to HU increases the likelihood that genetic studies using pediatric populations may reveal more precise basic biologic insights.
Genetic analysis of HU-induced HbF {#S7}
----------------------------------
Analyses of HbF regulation are crucial to understanding the spectrum of SCD severity, variability of HU response and design of novel therapies. In addition to the established observations of ethnic variability of HbF levels, several key observations drive the rationale for identifying genetic components of HU induction of HbF in U.S. populations of SCD:
1. Baseline HbF levels in SCD have high heritability^[@R55],[@R56]^;
2. HbF induction from HU therapy is also a heritable trait^[@R55]^;
3. Genome-wide SNP studies in normal non-anemic adults identified on a few major loci associated with variation of low HbF levels. These regions are both cis and trans to the beta globin gene locus^[@R56],[@R57]^;
4. These same loci are associated with baseline HbF in people with SCD in the U.S. ^[@R56],[@R58]--[@R62]^. Additional loci have been identified but not yet replicated;
5. A modest correlation in children exists between levels of HbF at baseline and on HU^[@R14],[@R33],[@R51]^.
Taken together, these findings lead to the prediction that genetic regulation of HbF expression at baseline overlaps with the control of HU-induced HbF The three major loci related to HbF expression in normal and SCD populations are: a SNP upstream of the gamma globin gene within the globin locus on chromosome 11, previously identified by restriction enzyme analysis as the XmnI site^[@R37],[@R60],[@R61]^; BCL11A, a transcription factor now recognized as a major silencer of HbF expression^[@R51],[@R58],[@R59],[@R61],[@R63]^; the intergenic interval between the HBS1L and MYB^[@R56],[@R58],[@R61]^. Additional loci have been identified and await replication^[@R59],[@R64],[@R65]^, in addition to epigenetic effects, as well as probable epigenetic effects ([Figure 3](#F3){ref-type="fig"})^[@R66]^.
Only a few published studies report on the genetics of HbF response to HU in SCD^[@R33],[@R51],[@R54],[@R64]^. Compared to genomic studies of more common disorders, sample sizes of HU effects on SCD are inevitably modest. Using the retrospective cohort from the MSH adult trial and assessing over two dozen candidate genes, Ma et al. reported significant associations between SNPs and HbF response to HU in loci of genes involved in the metabolism of arginine to nitric oxide and in a transcription factor that induces DNA bending. This report pre-dated the identification of BCL11A as a central regulator of HbF expression. Most of the MSH patients exhibited a small HbF response to HU^[@R54]^, with less than a 5% change in HbF from baseline. This blunted response is not universal in U.S. adults with SCD, and may be influenced by patient characteristics as well as adherence to HU regimen.
Whether HbF induction by HU occurs directly through via the direct influence of BCL11A, is a concept awaiting direct testing. BCL11A effects on HbF are probably mediated through its protein partners, upstream or downstream effectors, chromatin structure and/or telomerase function (recently reviewed)^[@R67]^. Other reports include associations between HU response in SCD and polymorphisms in the guanosine triphosphate (GTP)-binding protein gene sar1a^[@R64]^, underscoring the complexity of the genetic pathways regulating the HbF response to HU ([Figure 3](#F3){ref-type="fig"}).
Two pediatric pharmaco-genetics analyses using candidate single marker SNP polymorphisms suggested that just a few genes are associated with baseline HbF, including several SNPs within BCL11A ([Table 2](#T2){ref-type="table"}). SNP associations with induced HbF are generally not independent of baseline HbF levels^[@R33],[@R51]^. In contrast to the induced HbF level, the treatment-associated increment appears to be a less relevant marker. Both of these observations probably reflect the association between baseline and induced levels.
In our own, smaller multi-site analysis, baseline levels candidate were significantly associated with SNPs within the BCL11A and the beta and epsilon globin loci (HBB and HBE, respectively), with an additive attributable variance from these loci of 23% ([Table 2](#T2){ref-type="table"})^[@R51]^. Consistent with studies by Ware and colleagues^[@R14],[@R33]^, we reported that baseline HbF levels explained 33% of the variance in induced levels. The variant in HBE accounted for an additional 13% of the variance in induced levels, while variants in the HBB and BCL11A loci did not contribute beyond baseline levels. Thus our data suggest that the combined effects of baseline HbF and one SNP marker contributed an estimated 46% of the variance in HbF^[@R51]^.
By trend analysis, children with an allele associated with higher HbF ("favorable" allele) in one of the BCL11A and/or either globin marker had significantly higher average values of baseline HbF than those who lacked a favorable allele^[@R51]^. Effects on baseline HbF from a SNP in each these two genes were additive, and were associated with two-fold higher HbF for patients with favorable alleles in both loci. Similarly, having at least one favorable allele in either globin locus and BL11A was associated with a higher level of induced HbF. Statistical significance did not withstand adjustment for baseline fetal hemoglobin, likely reflecting the inter-relatedness of HbF regulation under both physiologic conditions. Genetic studies examining larger pediatric populations on HU, unusual responders and the influence of specific sequence variants are needed to the contribution of these and other genetic loci responsible for HbF response.
OTHER PHYSIOLOGIC EFFECTS OF HU {#S8}
===============================
Cellular biology {#S9}
----------------
The effects of HU largely depend upon its effects on nucleic acid synthesis in dividing red blood cell progenitors. HU affects the S-phase by inhibiting ribonucleotide reductase, an enzyme important for DNA synthesis. Depletion of DNA precursors by HU causes arrest of the replication fork, leading to cell death. A cell-based ex vivo assay for HbF induction, burst-forming unit erythroid (BFU-E) colonies grown in methylcellulose from blood of children with SCD, demonstrated that HU decreases the number of BFU-E colonies. HU and other ribonucleotide reductase inhibitors increase HbF production in that system. Interestingly, other cytotoxic agents that are not of that drug category, such as cytarabine and alkylating agents, decreased BFU-E counts but did not induce HbF^[@R41],[@R68]^.
HU's lethal effect of on ribonucleotide reductase and cell survival are also seen in laboratory bacteria such as E. coli^[@R69]^. Whether the bactericidal effects influence investigation using animal models or even patients has not previously been raised. Direct bacterial effects should be addressed on the HU-dampened expression of adhesion molecules in a murine model of bacterial infection.
HbF response to temporary marrow toxicity is probably attributable to transcriptional and epigenetic effects on the progenitor developmental program^[@R66],[@R70]^. HU signaling appears to involve cyclic guanine monophosphate (cGMP), cAMP, p38MAPK, and others pathways. Activation of cGMP may induce HbF via enhancing production of nitric oxide (NO)^[@R47],[@R68],[@R71],[@R72]^. NO may also support HbF production^[@R47]^. HU induces a small guanosine triphosphate (GTP)-binding protein, secretion-associated and RAS-related (SAR)^[@R73]^. SAR may be involved in the activation of transcription factors and signal transduction pathways in erythroleukemia K562 cells and in human bone marrow-derived progenitor cells. HU may also function through kinase and signal transduction pathways, such as GATA-1, to enhance gamma and beta globin synthesis in erythroid cells^[@R47]^.
Gene expression {#S10}
---------------
Comparing whole mRNA pre- and post- initiation of therapy revealed that HU affects expression of a number of genes involved in transcription and translation, such as ribosome assembly and chromosomal organization^[@R66],[@R70]^. Results may vary with age, dosing or other clinical conditions. Variation in cell source would be expected to affect detection of expressed genes, whether from bone marrow or purified early reticulocytes. HU may also affect expression of genes that link HU and HbF to BLCL11A^[@R66],[@R70]^. Epigenetic analysis of the gamma globin promoter did not find reveal much impact from HU^[@R74]^. Interestingly, HU appears to up-regulate specific miRNAs^[@R74]^. These results require further investigation, but underscore the view that HU is involved with complex pathways of gene regulation.
Testing for oncogenicity {#S11}
------------------------
The primary effect is damaging DNA replication by inhibiting ribonucleotide reductase. This effect raises concerns about an oncogenic potential, especially after prolonged use. These fears have been amplified by its original use as chemotherapy for chronic myeloid leukemia (CML), the latent phase of acute leukemia. While links acute leukemia outside of CML have been disproven^[@R75]^, concerns for the safety of long-term use in children persist. Several studies have tested DNA and cellular toxicity from pediatric HU users. No genotoxicity was detected using several different assays in vitro, including karyotype, illegitimate VDJ recombination, chromatid breaks^[@R9]^. Increased reticulocyte micronuclei were observed, but this effect was highly variable between patients and did not increase with time^[@R76]^. In all, oncogenicity of HU is probably quite low or non-existent. A few cases of acute leukemia were reported in patients after many years of HU treatment, but do not appear to be more frequent than in the untreated population^[@R75]^.
POTENTIAL PHARMACOLOGIC ALTERNATIVES TO HU {#S12}
==========================================
Other HbF inducers have been assessed over the past few decades, including nucleoside analogs such as 5-azacytidine and decitabine. However, they are often poorly tolerated potentially oncogenic and lack proof of effectiveness comparable to HU (recently reviewed)^[@R41]^. Additional HbF-inducing drugs are histone deacetylase inhibitors, erythropoietin (already high in SCD and shown not to induce HbF in SCD), valproate, thalidomide derivatives (e.g. pomalidimide), kit ligand. In all, a variety of cellular stresses and stimuli can promote coordinated stress responses, including activation of the gamma globin gene^[@R66],[@R70]^. Based on results from use of SCD mouse models, inhibitor of phosphodiesterase 9^[@R71]^ or inducible factor-1α (HIF-1α)^[@R77]^, alone or in combination with HU, may be clinically useful to stimulate cyclic GMP (cGMP) and nitric oxide for HbF production and/or to enhance its anti-sickling impact^[@R71]^.
BARRIERS TO HU UTILIZATION {#S13}
==========================
Outside of clinical trials with HU, ample documentation exists of incomplete clinical effectiveness of HU. Uneven drug adherence has been well documented^[@R78],[@R79]^. Provider non- and under-utilization is well documented^[@R75],[@R80]^. Our recent multi-site survey of parents of children with SCD revealed several family barriers to use of HU such as lack of FDA approval, near-universal safety concerns and highly varied knowledge about its benefits, including many for whom its basic property of decreasing episodes of pain was unknown^[@R81]^. Use of HU was positively correlated with fundamental knowledge of parents in the basic positive effects of HU on disease, independent of parental demographics such as education level, language spoken or ethnicity. Barriers in effective communication between providers and families may be exacerbated by issues arising from medical delivery systems.
The mixed uptake of HU by families may also reflect family perspectives on the long-term effects of SCD. A single site survey of parents revealed that the majority believed that the disease effects were going to diminish over time, and would not impact life goals or lifespan^[@R82]^. These poignant perceptions will need to be addressed if families are to embrace the long-term benefits of HU against its inconveniences and largely theoretical risks.
CONCLUSION {#S14}
==========
HU is a remarkably effective drug for a large proportion of children with SCD. Expanded understanding of the scientific underpinnings of its effects on SCD, the ability to predict individual response and the clinical applications for modifying disease effects are ongoing. Clinical trials will continue to test the uncertain benefits of HU ([Table 1](#T1){ref-type="table"}), such as primary prevention of brain infarcts (clinicaltrials.gov/show/NCT01389024). Murine models will facilitate insight into the benefits provided by induced HbF, altered expression of adhesion molecules, reduced BCL11A and other mechanisms. Genetic epidemiology will be used to identify specific variants in regulatory genes and gene pathways.
The accumulating science of HU is anticipated to lead to three direct effects for children with HbSS: 1) Use at earlier ages; 2) Wider clinical indications; and 3) Delineation of children who are less likely to enjoy substantive benefit from HU. For this last group, more aggressive consideration of chronic transfusion, hematopoietic stem cell transplantation or trial of emerging alternative agents may be warranted. To date, early clinical trials of other experimental HbF-inducing drugs have demonstrated considerable short- and long-term toxicity compared to HU. Therefore, HU is predicted to remain the mainstay of pharmacologic therapy for SCD in the foreseeable future.
Progress towards the Healthy People 2020 goals will occur through increased use of HU. Nonetheless, the entire SCD population may not benefit from HU alone. Dampened impact on clinical complications and HbF induction occurs for those with certain genotypes, many patients with HbSC, some adult patients and those with renal compromise. New, effective and safe therapies, alone or in combination with HU, are still needed to maximize pharmacologic benefit for everyone living with SCD. Constructive engagement must be made to assist families in undertaking long-term HU use to help them to balance the optimism of HU treatment and its potential toxicities with the risk of accumulating disease consequences. While several important crisis-modulators are currently under investigation, disease modifiers that prevent crises and other morbidities is arguably the primary therapeutic target.
**STATEMENT OF FINANCIAL SUPPORT**: The authors acknowledge support from the Clinical Translational Science Award (CTSA) at Columbia University, 5UL1RR024156 (H.N. Ginsberg, PI)
**DISCLOSURES:** No commercial or conflict of interest is reported by the authors.
{#F1}
{#F2}
{#F3}
######
Clinical Effects of Hydroxurea on Children with SCD[a](#TFN2){ref-type="table-fn"}
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Effects Blood/Circulation Organ/Whole body
---------------------------------------- ----------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------
**Laboratory/Physiologic measurement** Elevate HbF levels^[@R8],[@R14],[@R23]^ and stabilize high infant levels^[@R18]^\ Brain: Improve TCD flow velocity ^[@R26],[@R27]^\
Increase hemoglobin^[@R13]--[@R15]^\ Spleen: Preserve blood flow^[@R15],[@R18]^\
Increase MCV^[@R8],[@R14],[@R15],[@R22]^\ [b](#TFN3){ref-type="table-fn"} Lungs: Decrease acute chest syndrome^[@R15],[@R18]^\
Decrease hemolysis^[@R13]--[@R15]^\ Renal: Decrease hyperfiltration^[@R24]^ and hyposthenuria^[@R25]^
Decrease WBC and platelet counts^[@R8],[@R13]--[@R15]^
**Clinical/Well-being**\*\* [b](#TFN3){ref-type="table-fn"} Fewer acute pain crises^[@R8],[@R13],[@R18]^\ [b](#TFN3){ref-type="table-fn"},[c](#TFN4){ref-type="table-fn"} Reduce mortality^[@R30]--[@R32]^\
[b](#TFN3){ref-type="table-fn"} Reduce dactylitis^[@R18]^\ [b](#TFN3){ref-type="table-fn"} Improve growth^[@R13],[@R15]^\
[b](#TFN3){ref-type="table-fn"} Fewer transfusions^[@R18]^\ Improve quality of life^[@R21],[@R22]^
[b](#TFN3){ref-type="table-fn"} Fewer hospitalizations^[@R8],[@R13],[@R18]^
**Not yet known** Stabilize HbF as adults\ [b](#TFN3){ref-type="table-fn"} Improve overall lifespan for children\
Reduce allo-immunization (through reducing transfusion)\ [b](#TFN3){ref-type="table-fn"} Improve cognitive development\
Reduce transfusion-related iron toxicity (through reducing transfusion) Protect from stroke/infarct\
Prevent long-term renal, lung, cardiac effects\
Reduce cholelithiasis\
Reduce retinopathy\
[b](#TFN3){ref-type="table-fn"} Normalize timing of physical maturation\
[b](#TFN3){ref-type="table-fn"} Maintain fertility
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
MCV -- Mean red cell volume TCD -- Transcranial Doppler
Some of the effects have not been demonstrated across all pediatric age ranges nor by prospective randomized trials. Reports demonstrating effect by randomized trials are preferentially cited.
Patient-oriented outcomes
Established long-term outcome for adults
######
SNP Polymorphisms Associated with HbF
*Gene* Phenotype Clinic Population Ethnicity N SNP β p Ref. \#
------------- ------------------- ------------------- ------------ ----------- ------ ------------ ------- ------------- ---------
*ARG1* HbF induced by HU pediatric USA AA 174 rs17599586 NA 4 × 10^−3^ [@R33]
*ARG2* HbF induced by HU pediatric USA AA 174 rs2295644 NA 3 × 10^−3^ [@R33]
*ARG2* HbF induced by HU adult USA AA 137 rs10483801 NA 1 × 10^−3^ [@R54]
*BCL11A* HbF baseline pediatric USA AA, H 108 rs4671393 2.88 5 × 10^−5^ [@R51]
*BCL11A* HbF baseline pediatric USA AA 174 rs4671393 NA 3 × 10^−4^ [@R33]
*BCL11A* HbF baseline adult USA AA 255 rs766432 NA 2 × 10^−10^ [@R63]
*BCL11A* HbF baseline pediatric + adult USA AA 1032 rs4671393 0.60 4 × 10^−37^ [@R60]
*BCL11A* HbF baseline pediatric + adult USA AA 1275 rs4671393 0.60 2 × 10^−42^ [@R61]
*BCL11A* HbF baseline pediatric + adult Brazil NA 350 rs4671393 0.50 3 × 10^−8^ [@R61]
*BCL11A* HbF induced by HU pediatric USA AA, H 47 rs1186868 3.37 0.019 [@R51]
*FTL1* HbF induced by HU adult USA AA 137 rs2182008 NA 0.003 [@R54]
*GLP2R* F cells baseline pediatric + adult USA AA 440 rs12103880 −1.36 3 × 10^−8^ [@R59]
*HAO2* HbF induced by HU adult USA AA 137 rs10494225 NA 2 × 10^−3^ [@R54]
*HBB* HbF baseline pediatric USA AA, H 108 rs7482144 3.88 2 × 10^−4^ [@R51]
*HBB* HbF baseline pediatric + adult USA AA 1032 rs10128556 0.42 2 × 10^−9^ [@R60]
*HBB* HbF baseline pediatric + adult USA AA 1275 rs7482144 0.41 4 × 10^−7^ [@R61]
*HBE* HbF baseline pediatric USA AA, H 108 rs7130110 2.86 6 × 10^−5^ [@R51]
*HBE* HbF baseline pediatric USA AA 174 rs7130110 NA 3 × 10^−5^ [@R33]
*HBE* HbF induced by HU pediatric USA AA, H 38 rs7130110 6.04 0.004 [@R51]
*HBS1L-MYB* HbF baseline pediatric + adult USA AA 1032 rs9402686 0.65 2 × 10^−13^ [@R60]
*HBS1L-MYB* HbF baseline pediatric + adult USA AA 1275 rs9399137 0.60 5 × 10^−11^ [@R61]
*NOS1* HbF induced by HU adult USA AA 137 rs7977109 NA 0.023 [@R54]
*OR51B5/B6* HbF induced by HU pediatric USA AA 1153 rs5006884 0.20 3 × 10^−8^ [@R65]
*OR51B6* HbF baseline pediatric USA AA, H 108 rs5024042 1.70 0.031 [@R51]
*SAR1A* HbF induced by HU adult USA AA 32 rs4282891 NA \< 0.05 [@R64]
AA: African American; H: Hispanics; AC: African Caribbean; WA: West African; N: sample size; All: phenotype associated allele
| {
"pile_set_name": "PubMed Central"
} |
Oropharyngeal dysphagia is a deglutition disorder resulting from neurological and/or structural impairment that may result in in decreased social interaction, loss of pleasure in food consumption, malnutrition, dehydration, infectious pulmonary processes and other high morbidity processes. Patients with heart disease from various causes present a risk of developing oropharyngeal dysphagia.
Heart patients are susceptible to neurological disorders such as stroke and transient ischemic attack that may be present with dysphagia. Other possible risk factors for oropharyngeal dysphagia in people with heart disease are: prolonged tracheal intubation, dependence on mechanical ventilation, presence of a tracheostomy tube, manipulation or compression of cranial nerve pairs involved in swallowing and use of drugs that increase dysphagia, in addition to the physiological changes common in elderly patients, especially those with comorbidities, that may interfere in the different phases of deglutition and contribute to the emergence of dysphagic symptoms^[@r01],[@r02]^.
The speech therapist is the most qualified professional to assess and rehabilitate patients with dysphagia. Once dysphagia is diagnosed, a therapeutic program must be designed to regain efficient swallowing without the risk of pulmonary, nutritional and psychological complications.
Promoting safe swallowing for individuals with dysphagia is a challenge that can be facilitated using therapeutic resources such as diet adaptation, with changes in consistency, volume, temperature and flavor. These strategies used by speech therapists are part of swallowing rehabilitation because they interfere with sensorimotor oral performance and oropharyngeal transit, minimizing the risk of *laryngotracheal* aspiration^[@r03]^.
A change in the consistency of food is an important tool in the treatment of dysphagia, and foods should be modified according to the degree of dysphagia, the patient\'s nutritional condition, food intake acceptance and patient morbidity.
Food thickener is used for liquid thickening, a common practice in speech therapy, since liquid swallowing requires greater control than swallowing of foods with a thicker consistency. The chance of premature spillage of food is one of the biggest risks of penetration and/or *larygotracheal* aspiration^[@r04]^.
According to the standardization of the National Dysphagia Diet (NDD)^[@r05]^, liquids may be thickened into three consistencies: nectar, honey and pudding, which are prescribed by the speech therapist according to the degree of the patient\'s dysphagia. The viscosity of thickened liquids is measured in centipoises (cP) and is categorized according to the NDD as: thin liquid 1-50 cP, nectar 51-350 cP, honey 351-1750 cP and higher, and pudding 1750 cP.
The amount of thickener indicated for each consistency will depend on the makeup of the different brands available on the market.
Nutrition (dietary) and medical teams are responsible for the patient\'s diet in order to maintain or restore the patient\'s nutritional status. The heart patient often requires sodium restriction, which must be achieved with a maximum amount of 5 g of sodium chloride per day, which is the value of the entire daily intake. The different restriction levels are guided by the patient\'s individual pathology and the clinical path of each patient. Sodium restriction may be mild, moderate or severe; in the case of moderate restriction, intake should be 2g of salt per day and in the case of severe restriction the patient should eat no added salt, only the salt that is intrinsic in their food^[@r06],[@r07]^.
Thus, it is necessary in daily interaction with patients to pay attention to the nutritional information of the adapted diet, especially in regards to thickened liquids and the heart patient. This quality standardization is achieved by the presence of mineral sodium in the composition of the thickeners.
The amount of sodium present in the thickeners is not uniform or standardized, varying according to brands available on the market ([Table 1](#t01){ref-type="table"}).
######
Sodium amount in commercialized thickeners
**Thickeners** **Sodium per serving**
---------------- ------------------------
Brand 1 8mg in 4.5g
Brand 2 13mg in 1,2g
Brand 3 2,4mg in 5.0g
Brand 4 0mg in 6.0g
Brand 5 0mg in 3.0g
Brand 6 4.6 mg in 3 g
Brand 7 0mg in 5g
Brands - 1: Thick & Easy; 2: Resource Thicken Up Clear; 3: Biosen; 4:Sust\'up; 5: MaxiSperse; 6: Nutilis; 7: Nutriclin.
The total daily intake of sodium in the thickened liquid will depend on the consistency indicated by the speech therapist, the volume of liquid prescribed by the medical team and the brand of the thickener used. [Table 1](#t01){ref-type="table"} shows samples of different brands of thickener and the amount of sodium in the three consistencies: nectar, honey and pudding, for a daily intake of 1500 ml of liquids.
Despite the amount of daily sodium intake in the supply of the thickened liquid at the different consistencies, this amount is not high. However, according to the samples in [Table 2](#t02){ref-type="table"}, this amount could compromise the total sodium intake in moderate or severely restricted diets. Thus, this value must always be considered in the overall diet, taking into account the minimum restriction of 5g sodium chloride; that is, 2,000 mg sodium/day^[@r06]^.
######
Daily sodium amount (mg) in thickener in three consistencies: nectar, honey and pudding: (1500 ml of liquid)
**Consistencies** **Sodium amount (brand 1)** **Sodium amount (brand 2)** **Sodium amount (brand 3)**
------------------- ----------------------------- ----------------------------- -----------------------------
Nectar 195 mg 137 mg 120 mg
Honey 390 mg 171 mg 180 mg
Pudding 585 mg 220 mg 240 mg
Brands: 1: Thicken Up Clear/Nestle; 2: Nutilis/Support; 3: Thick & Easy/Fresenius.
It is important that the speech therapist, in partnership with medical staff and dietary staff, carefully evaluate the patients who are prescribed thickened liquids, considering that the value of sodium added by the use of thickener in the general diet is liable to impact negatively on the patient\'s clinical condition.
In cases with significant restrictions of sodium intake, the intake related to the thickener can be a significant impact factor. It is essential that the team involved considers this aspect towards better therapeutic management of the case.
An opportunity to discontinue the thickener should be evaluated, and the role of the speech therapist is to identify if the patient is able to discontinue use of the thickener, if he will be using it for a long time or even continuously, and to inform the medical staff who will check into the impact in terms of sodium in each case, correlating it with the time of use.
Another aspect that should be decided as a team is the possibility of using natural thickeners, such as the use of fruit that provides a thicker consistency in the preparation of juices and vitamins; in addition, it is also possible for the nutrition team to adapt a diet that avoids foods rich in sodium, as well as a reduction in the amount of salt used in the preparation of food, thus reducing the value of the added sodium.
Some thickeners show an absence of sodium on their labels and the team may opt for these brands. However, there are many controversies regarding the quality of these thickeners, since sodium participates actively in the process of thickening.
The amount of sodium added to the thickener should not be a factor considered separately, but should be considered in the total sodium intake throughout the day. The team needs to observe the patient using thickener closely and be aware of the need to analyze each case separately, taking into account aspects such as basic disease and clinical condition.
The optimal treatment for the patient with dysphagia does not aim at guaranteeing the safe intake of food, but also an optimal nutritional and water intake, in addition to consequent improvement of the clinical condition of the patient. For this to occur, it is necessary to have the cooperation of the entire team, aimed at quality care.
Author contributions
====================
Conception and design of the research: Almeida TM, Germini MFC; Acquisition of data: Kovacs C, Soares AMNGF; Analysis and interpretation of the dataand Writing of the manuscript: Almeida TM; Critical revision of the manuscript for intellectual content: Magnoni D; Institutional support: Sousa AGMR.
**Potential Conflict of Interest:** No potential conflict of interest relevant to this article was reported.
**Sources of Funding:** There were no external funding sources for this study.
**Study Association:** This study is not associated with any post-graduation program.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s0001}
============
ABP 215 (US: MVASI™ \[bevacizumab-awwb\] approved in September 2017; EU: MVASI™ \[bevacizumab\] approved in January 2018; Amgen Inc.) is a biosimilar to bevacizumab (Avastin®), a recombinant humanized monoclonal immunoglobulin G1 (IgG1) antibody. Bevacizumab binds to vascular endothelial growth factor A (VEGF-A) and prevents the binding of VEGF-A to VEGF receptors on the surface of endothelial cells, inhibiting endothelial cell proliferation and new blood vessel formation, thereby leading to normalization of the tumor vasculature.[@cit0001] Bevacizumab was first approved by the United States Food and Drug Administration (FDA) in 2004, and then by the European Medicines Agency (EMA) in 2005, to treat patients with certain types of cancers where tumor vasculature contributes to tumor growth.[@cit0002]
ABP 215 is the first approved biosimilar to bevacizumab in the US and EU. ABP 215 is approved in the US for the treatment of metastatic colorectal cancer, non-squamous non--small cell lung cancer, glioblastoma, metastatic renal cell carcinoma, and persistent, recurrent, or metastatic carcinoma of the cervix.[@cit0003] In the EU, ABP 215 is approved for the treatment of metastatic carcinoma of the colon or rectum, metastatic breast cancer, non-squamous non--small cell lung cancer, metastatic renal cell cancer, ovarian cancer, fallopian tube or primary peritoneal cancer, and persistent, recurrent, or metastatic carcinoma of the cervix.[@cit0004]
Biosimilars have the potential to bring increased access to important therapeutics to a broader patient population. However, transforming complex therapeutic proteins into effective biological medicines requires highly specialized knowledge and experience with scientific standards, processes, and quality systems. Since each biosimilar product must establish a unique cell line and manufacturing process, biosimilars are not expected to be identical to their reference products. Instead, biosimilars will have minor differences in product attributes, which do not impact clinical safety and efficacy. The FDA and EMA provide guidelines for generation of a robust data package of analytical, biological, non-clinical and clinical data to demonstrate similarity of structure and function with no notable risks to the safety and efficacy of the product in clinical use.[@cit0005]
Here, we present the plan and results of a comprehensive analytical similarity assessment of Amgen\'s biosimilar ABP 215 to assess its analytical similarity with bevacizumab. The results also include a comparison of bevacizumab from the US (bevacizumab \[US\]) and European Union (bevacizumab \[EU\]). The analytical similarity assessment plan was designed to assess structural/physicochemical and functional similarity and ensure the understanding of whether any differences between ABP 215 and bevacizumab had the potential to impact clinical performance, consistent with US and EU regulatory guidelines.[@cit0005]
Results {#s0002}
=======
The testing plan listing all the analytical techniques and attributes/assays evaluated for ABP 215 and bevacizumab is shown in [Table 1](#t0001){ref-type="table"}. The purpose of the plan was to evaluate both active ingredients and inactive ingredients that could affect product safety and efficacy in addition to product quality. Where applicable, orthogonal methods were used to fully analyze product attributes and activities. Table 1.Similarity testing plan and the analytical methods for the structural and functional characterization of the proposed biosimilar ABP 215 and bevacizumab reference products.CategoryAnalytical TechniquePrimary StructureMolecular mass of intact whole protein Molecular mass of reduced and deglycosylated HC and LC Protein sequence by reduced peptide map Disulfide structure by non-reduced peptide map N-glycan map by HILIC HPLC Extinction coefficient by amino acid analysis Isoelectric point by cIEF Identity by anti-idiotype ELISAHigher Order StructureSecondary structure by FTIR Tertiary structure by near UV-CD Thermal stability by DSCParticles and AggregatesSubvisible particles by light obscuration Subvisible particles by MFI Submicron particle profile by DLS Submicron particle profile by FFF Aggregates by AUC-SV Aggregates by SE-HPLC-LSProduct-related Substances and ImpuritiesSize variants by SE-HPLC, rCE-SDS and nrCE-SDS Charge variants by CEX-HPLCThermal Forced DegradationThermal stability at 25, 40, and 50°C assessed by purity and potencyBiological ActivityVEGF-A binding (ELISA) Proliferation inhibition bioassay (potency) VEGF-A binding kinetics and affinity Binding to VEGF-A isoforms Inhibition of VEGFR-2 RTK autophosphorylation Specificity by VEGFR-2 RTK autophosphorylation FcRn binding FcγRIa binding FcγRIIa binding FcγRIIb binding FcγRIIIa (158 V) binding FcγRIIIa (158 F) binding FcγRIIIb binding C1q binding Lack of ADCC activity Lack of CDC activityGeneral PropertiesProtein concentration and volume Osmolality, pH, appearance, color, and clarityProcess-Related ImpuritiesHCP by ELISA, 2D LC-MS, and 2D-DIGE Residual protein A by ELISA Residual DNA analysis by qPCR[^1]
More than 20 batches of US bevacizumab and EU bevacizumab were tested over six years. The ABP 215 batches used in similarity assessment represent drug product lots filled from 13 individual drug substance lots manufactured over the same period. At least 13 ABP 215 lots were tested for product attributes such as purity and glycan structure, which could be affected by the drug substance manufacturing process. For product attributes that could be influenced by fill--and-finish process, including volume and particles, all drug product batches were tested for similarity. This testing plan resulted in a large number of datasets, which enabled meaningful comparisons, including the use of statistical analysis per regulatory agency guidance, and helped further garner confidence in the overall conclusion for the analytical similarity assessment of ABP 215 to both US- and EU-sourced bevacizumab.
Primary structure {#s0002-0001}
-----------------
To investigate the analytical similarity of the primary structures of ABP 215 and bevacizumab, several complementary characterization methods were used, employing state-of-the-art analytical tools, such as high precision and accuracy mass spectrometry for protein sequence, and high-resolution chromatography with sensitive detection for oligosaccharide profiles.
Results of intact mass analysis, reduced peptide map, and glycan map are shown in [Fig. 1a-c](#f0001){ref-type="fig"}, respectively. As shown in [Fig. 1a](#f0001){ref-type="fig"}, the predominant species are consistent with the presence of two core-fucosylated complex N-linked glycans with either 0, 1, or 2 terminal galactose residues, depicted as A2G0F, A2G1F, and A2G2F, respectively. Peaks A, B, C, D, and E were identified as the predominant forms with molecular weights consistent with a structure that contains two glycans composed of A2G0F:A1G0F (peak A); A2G0F:A2G0F (peak B); A2G0F:A2G1F (peak C); A2G1F:A2G1F or A2G0F:A2G2F (peak D); and A2G1F:A2G2F (peak E). The intact molecular mass of all predominant species matched the theoretical mass for the products ([Table 2](#t0002){ref-type="table"}). Figure 1.Primary Structure: Comparison of ABP 215, Bevacizumab (US), and Bevacizumab (EU) Intact Molecular Mass Profile (a), reduced peptide map (b), and glycan map (c). Table 2.Summary of attributes for ABP 215, bevacizumab (US), and bevacizumab (EU).Analytical Testing/AttributesABP 215 Range (n)Bevacizumab US Range (n)Bevacizumab EU Range (n)Intact molecular weight (Da) A: Glycosylation -- A2G0F:A1G0F149008--149017 (4)149007--149015 (4)149009--149025 (4) B: Glycosylation -- A2G0F:A2G0F149210--149219 (4)149210--149218 (4)149209--149218 (4) C: Glycosylation -- A2G0F:A2G1F149371--149381 (4)149374--149378 (4)149373--149381 (4) D: Glycosylation -- A2G1F:A2G1F or A2G0F:A2G2F149536--149541 (4)149531--149535 (4)149531--149538 (4) E: Glycosylation -- A2G1F:A2G2F149694--149698 (4)149673--149698 (4)149663--149710 (4) Reduced and deglycosylated heavy chain (Da)49744--49752 (4)49741--49753 (4)49742--49753 (4) Reduced and deglycosylated light chain (Da)23479--23482 (4)23479--23481 (4)23479--23482 (4) Glycosylation at Asn~303~ (%)99.2--99.4 (13)97.9--98.3 (24)97.9--98.4 (26)Glycan map (%) High mannose1.2--2.7 (13)0.3--1.3 (23)0.5--1.2 (25) Afucosylation1.2--1.7 (13)1.7--3.8 (23)1.9--2.6 (25) Galactosylation17.1--29.4 (13)8.7--21.8 (23)7.8--21.2 (25) Sialic acid0.2--0.3 (13)0.1--0.2 (23)0.1--0.2 (25)FTIR/spectral similarity (%) US RP97.6--99.8 (6)97.9--99.3 (6)97.2--99.0 (6) EU RP97.9--99.8 (6)98.3--99.7 (6)97.5--99.3 (6)Near UV-CD/spectral similarity (%) US RP98.5--99.4 (6)98.5--99.6 (6)97.3--99.4 (6) EU RP97.1--99.4 (6)97.0--99.6 (6)97.6--99.5 (6)DSC (°C) T~m1~72.7--73.1 (11)72.6--73.1 (12)72.5--73.1 (12) T~m2~83.2--83.7 (11)83.2--83.9 (12)83.0--84.0 (12)LO subvisible particle (particles/container) ≥ 2 µm320--5736 (19)48--4749 (14)77--15493 (15) ≥ 5 µm48--1488 (19)14--755 (14)24--688 (15) ≥ 10 µm16--219 (19)3--112 (14)8--224 (15) ≥ 25 µm0--96 (19)0--32 (14)0--32 (15)MFI subvisible particle (particles/mL) ≥ 5 µm spherical particles16--218 (19)42--3439 (11)51--588 (12) ≥ 5 µm non-spherical particles0--65 (19)6--2157 (11)8--430 (12) AUC-SV/monomer (%)98.0--99.3 (6)97.3--98.6 (6)97.3--98.6 (6)SE-HPLC-LS MW (kDa) Monomer148 (3)148 (3)148 (3) Dimer297--304 (3)300--301 (3)294--298 (3) Protein concentration (mg/mL)24.4--25.8 (13)23.9--25.9 (24)24.5--25.8 (26) Volume (mL) 100 mg/4 mL4.1--4.3 (7)4.2--4.3 (6)4.3--4.7 (5) 400 mg/16 mL16.3--16.5 (12)16.2--16.6 (6)16.3--16.8 (8) CHO cell protein by ELISA (ng/mg)4--12 (13)6 (3)5--6 (3)[^2]
The molecular masses of reduced and deglycosylated drug product provided further evidence that the heavy chain (HC) and light chain (LC) polypeptide compositions were similar between ABP 215 and bevacizumab ([Table 2](#t0002){ref-type="table"}). The observed masses for both ABP 215 and bevacizumab match closely to the theoretical masses calculated based on the coding oligonucleotide sequences, including the conversion of Asn~303~ to Asp~303~ as a result of removal of the N-glycans by PNGase F and the absence of the C-terminal lysine on the HC. The major peaks correspond to the expected theoretical masses of reduced and deglycosylated HC and LC. In addition, a minor peak with a mass addition of 162 Daltons corresponding to glycated HC and LC were present in both ABP 215 and bevacizumab. Glycation is a type of chemical modification, which takes place primarily during the cell culture process, and most commonly affects the *e*-amino side chain of lysine residues and N-terminal amines of a protein.[@cit0008] Protein glycation occurs under physiological conditions and the typical low levels of glycation found in monoclonal antibodies are not expected to have any safety impacts.[@cit0008] The levels of glycation observed in both HC and LC for ABP 215 and bevacizumab were all low and comparable.
The reduced peptide map overlays ([Fig. 1b](#f0001){ref-type="fig"}) showed similar peak profiles for ABP 215 and bevacizumab, and the differences between the observed and theoretical masses for all tryptic peptides were within the specified mass accuracy based on instrument and method capability. The same post-translational modifications were detected in both ABP 215 and bevacizumab, and no new species were detected. Examination of the amino acid sequence revealed a single consensus site for N-linked glycosylation located on the HC at Asn~303~ in all ABP 215 and bevacizumab peptide maps. LC-MS data showed the prevalent glycan structures to be core fucosylated biantennary complex structures with 0 or 1 terminal galactose, consistent with the glycan map results, as discussed in this section.
Several low abundance peptides were observed in both ABP 215 and bevacizumab. Characterization by mass spectrometry identified the modifications, which included non-glycosylation at Asn~303~ site, cyclization of the HC N-terminal Glu~1~ residue to pyroglutamate (pE), truncation of the HC C-terminal Lys~453~ residue, and a few deamidation and oxidation sites. The same modifications and sites were observed in both ABP 215 and bevacizumab, and the levels of these post-translational modifications were similar.
In addition, the non-reduced peptide map confirmed that both ABP 215 and bevacizumab contain a total of 32 cysteine residues comprising 11 cysteines in each HC and 5 cysteines in each LC. Under native conditions, these residues form a total of 16 disulfide bonds, including 12 intrachain and 4 interchain disulfide bonds. The presence of the expected 16 cysteine-containing residues was confirmed for both ABP 215 and bevacizumab. The difference between the observed molecular masses and the theoretical masses for the disulfide-linked peptides in the non-reduced peptide maps, and the cysteine-containing peptides in the reduced peptide maps, were all within the specified mass accuracy based on instrument and method capability.
The overlay of glycan maps for ABP 215 and bevacizumab (US and EU) is shown in [Fig. 1c](#f0001){ref-type="fig"}. The results demonstrate ABP 215 and bevacizumab have similar profiles with a total of 19 glycan forms, and all major glycans are present in both ABP 215 and bevacizumab. As shown in [Table 2](#t0002){ref-type="table"}, four glycan groups were evaluated as part of the similarity assessment based on their potential to affect pharmacokinetics (PK) or biological functions, including binding to crystallizable fragment (Fc) receptors. The minor differences at these observed levels do not affect any of the biological activities, as shown in [Table 3](#t0003){ref-type="table"}. Table 3.Summary of functional attributes for ABP 215, bevacizumab (US), and bevacizumab (EU).Analytical Testing/AttributesABP 215 Range (n)Bevacizumab US Range (n)Bevacizumab EU Range (n)Inhibition of proliferation (potency) (%)91--105 (13)86--104 (24)88--103 (25)Relative VEGF-A binding (%)84--98 (13)80--104 (14)76--99 (13)VEGF-A binding affinity (K~D~)20.1--23.5 (3)22.9--23.9 (3)18.4--23.4 (3)Binding to FcRn (%)84--104 (13)91--107 (14)88--103 (14)Binding to FcγRIa (%)90--98 (3)83--90 (3)99--102 (3)Binding to FcγRIIa (%)83--101 (3)89--91 (3)80--90 (3)Binding to FcγRIIb (%)92--98 (3)90--94 (3)87--91 (3)Binding to FcγRIIIa, 158V (%)78--115 (13)77--98 (12)81--117 (17)Binding to FcγRIIIa, 158F (%)84--120 (13)80--110 (10)83--109 (10)Binding to FcγRIIIb (%)89--92 (3)80--85 (3)67--88 (3)Binding to C1q (%)91--122 (13)88--115 (10)90--104 (9)[^3]
High-mannose glycan forms, including M5, M6, M7, and M8, have the potential to affect PK via differential clearance through binding to mannose-binding receptors.[@cit0009] In addition, they could influence antibody-dependent cell-mediated cytotoxicity (ADCC) via changes in binding to FcγRIIIa.[@cit0010] However, due to the soluble nature of the target, ABP 215 and bevacizumab do not exhibit ADCC activity.[@cit0011] ABP 215 had slightly higher levels of high mannose ([Table 2](#t0002){ref-type="table"}), but the minor differences at the levels observed have no effect on FcγRIIIa binding activity ([Table 3](#t0003){ref-type="table"}) and are not considered clinically meaningful since ABP 215 has a similar PK profile compared with bevacizumab.[@cit0009]^,^[@cit0012]
Afucosylated glycan forms, N-linked glycans that lack core fucose, include complex-type, hybrid-type, and high mannose glycans.[@cit0012] Afucosylation of the N-linked glycan of the Fc region can enhance FcγRIIIa binding of IgG1 antibodies.[@cit0010] ABP 215 had slightly lower levels of afucosylation ([Table 2](#t0002){ref-type="table"}), but this minor difference at the level observed has no impact on FcγRIIIa binding activity ([Table 3](#t0003){ref-type="table"}) and is not considered clinically meaningful since ABP 215 and bevacizumab do not exhibit ADCC activity.[@cit0011]
Galactosylated glycan forms include all complex and hybrid glycan structures that contain at least one terminal galactose.[@cit0012] Galactosylation has the potential to affect binding to C1q, which is a key step in the activation of IgG1-mediated cell lysis via complement-dependent cytotoxicity (CDC).[@cit0013]^,^[@cit0014] ABP 215 had slightly higher levels of galactosylation ([Table 2](#t0002){ref-type="table"}), but the minor difference at the level observed has no impact on C1q binding activity ([Table 3](#t0003){ref-type="table"}) and is not considered clinically meaningful since ABP 215 and bevacizumab do not exhibit CDC due to the soluble nature of the target.[@cit0011]
Sialylated glycan forms include all complex and hybrid glycan structures that contain at least one terminal sialic acid.[@cit0015] Both ABP 215 and bevacizumab have similarly low levels of sialyation at or near the limit of quantitation (0.1%) of the assay ([Table 2](#t0002){ref-type="table"}), which have no biological impact as demonstrated by the results of the functional similarity assessment.
The apparent isoelectric point (pI) of the protein is influenced by the amino acid sequence and its higher order structure. The apparent pI evaluated using capillary isoelectric focusing (cIEF) showed that the charge distribution profiles for ABP 215 and bevacizumab were similar, and no new peaks were observed ([Fig. 2d](#f0002){ref-type="fig"}). Figure 2.Physicochemical Properties of Size and Charge Variants, as Assessed by SE-HPLC (a), rCE-SDS (b), CEX-HPLC (c), and cIEF (d) for ABP 215 (□), bevacizumab (US) (∆), and bevacizumab (EU) (○).
Higher order structure {#s0002-0002}
----------------------
The higher order structure of ABP 215 and bevacizumab was characterized using multiple biophysical techniques.
Results of secondary structure by Fourier-transform infrared spectroscopy (FTIR), tertiary structure by ultraviolet circular dichroism (UV-CD), and thermal stability by differential scanning calorimetry (DSC) are shown in [Fig. 3a-c](#f0003){ref-type="fig"}, respectively. The FTIR profiles were visually similar between ABP 215 and bevacizumab ([Fig. 3a](#f0003){ref-type="fig"}). The spectra exhibit strong β-sheet bands at around 1639 cm^−1^ and at 1689 cm^−1^, indicating the presence of antiparallel β-sheet structure existing in typical antibodies. The spectral similarity values of ABP 215 and bevacizumab are ≥ 95% ([Table 2](#t0002){ref-type="table"}), with 95% being the limit of the method\'s precision. The secondary structures of ABP 215 and bevacizumab were demonstrated to be similar. Figure 3.Higher Order Structure: Comparison of ABP 215, Bevacizumab (US), and Bevacizumab (EU) Fourier-transform infrared spectra (a), ultraviolet circular dichroism spectra (b), and differential scanning calorimetry thermograms (c).
The near UV-CD profiles were visually similar between ABP 215 and bevacizumab ([Fig. 3b](#f0003){ref-type="fig"}). The spectra contain signals from tryptophan, tyrosine, and phenylalanine, superimposed on the broad disulfide signal from 250 to 280 nm, indicating the presence of native tertiary structure and that the disulfide bonds and aromatic amino acids are in the expected environment due to proper folding of the proteins. The spectral similarity values of ABP 215 and bevacizumab are ≥ 95% ([Table 2](#t0002){ref-type="table"}), with 95% being the limit of the method\'s precision. The tertiary structure of ABP 215 and bevacizumab is demonstrated to be similar.
When visually compared, the DSC profiles of ABP 215 and bevacizumab were similar ([Fig. 3c](#f0003){ref-type="fig"}). The DSC profiles have two endothermic thermal transitions corresponding to the unfolding of the antigen-binding fragment (Fab)/C~H~2 and C~H~3 domains, as characterized by the DSC thermal melting temperatures (T~m~). The T~m1~ and T~m2~ values of ABP 215 and bevacizumab were similar, as summarized in [Table 2](#t0002){ref-type="table"}. The thermal stabilities of ABP 215 and bevacizumab were demonstrated to be similar.
Particles and aggregates {#s0002-0003}
------------------------
ABP 215 has similarly low levels of subvisible particles compared with bevacizumab as assessed by light obstruction (LO) and micro-flow imaging (MFI) ([Table 2](#t0002){ref-type="table"}). No submicron particles above the sensitivity of the assay were detected for any of the products as assessed by field flow fractionation (FFF) and dynamic light scattering ([Fig. 4a](#f0004){ref-type="fig"} and [4b](#f0004){ref-type="fig"}). Aggregates assessed by AUC-SV ([Fig. 4c](#f0004){ref-type="fig"} and [Table 2](#t0002){ref-type="table"}) and SE-HPLC-LS ([Table 2](#t0002){ref-type="table"}) were similar between ABP 215 and bevacizumab. These results demonstrate that ABP 215 has similar particle and aggregate levels compared with bevacizumab. Figure 4.Particles and Aggregates: Comparison of ABP 215, Bevacizumab (US), and Bevacizumab (EU) submicron particle profiles by Field Flow Fractionation (a), dynamic light scattering (b), and aggregates by analytical ultracentrifugation-sedimentation velocity (c).
Product-related substances and impurities {#s0002-0004}
-----------------------------------------
Product-related substances and impurities of ABP 215, bevacizumab (US), and bevacizumab (EU) were assessed using a combination of methods that evaluate size and charge variants. ABP 215 has a slightly lower level of aggregates, as shown in SE-HPLC ([Fig. 2a](#f0002){ref-type="fig"}), and non-glycosylated HC, as shown in reduced CE-SDS ([Fig. 2b](#f0002){ref-type="fig"}), compared with bevacizumab.
Assessment of charge variants by CEX-HPLC and cIEF showed that ABP 215 has a slightly higher level of basic variants due to unprocessed HC C-terminal lysine compared with bevacizumab. This was further confirmed by treatment with carboxypeptidase B (CpB), which showed the basic peaks were reduced to similar levels after CpB treatment for all three products ([Fig. 2c](#f0002){ref-type="fig"}). cIEF data supported the conclusion that the same charged species are present in all three products ([Fig. 2d](#f0002){ref-type="fig"}), with higher level of basic peak corresponding to unprocessed HC C-terminal lysine. Further characterization of the charge variants by peptide mapping showed the same modifications present in both ABP 215 and bevacizumab, and the minor quantitative differences observed does not affect biological activity ([Table 3](#t0003){ref-type="table"}).
Size and charge variants are stability-indicating attributes, which may change as a function of storage time at the recommended storage condition. Therefore, any similarity assessment of the quantitative levels of these product-related variants requires consideration of material age at the time of analysis. The results for the product-related variants were plotted against the estimated material age at the time of testing. ABP 215 and bevacizumab have similar degradation rates for the size and charge variants, as shown in [Fig. 2a](#f0002){ref-type="fig"} for SE-HPLC main peak, [Fig. 2b](#f0002){ref-type="fig"} for reduced CE-SDS purity (HC plus LC), and [Fig. 2c](#f0002){ref-type="fig"} for CEX-HPLC main peak.
Stability and degradation at accelerated conditions {#s0002-0005}
---------------------------------------------------
As part of the analytical similarity assessment, thermal stability and degradation studies were performed at 25°C, 40°C, and 50°C to aid in the comparison of product degradation pathways, which could be influenced by residual impurity from manufacturing process. Changes in the stability-indicating attributes were assessed using SE-HPLC, reduced CE-SDS, CEX-HPLC, and potency methods. Degradation profile results for thermally degraded test samples after incubation at 50°C for 14 days were shown to be similar between ABP 215 and bevacizumab for both the presence and the absence of degraded species, as illustrated for SE-HPLC ([Fig. 5a](#f0005){ref-type="fig"}). Similarity was further demonstrated by plotting degradation rates for all samples\' stability-indicating attributes at the designated intervals, as shown for SE-HPLC main peak at 50°C for up to 14 days ([Fig. 5b](#f0005){ref-type="fig"}). The degradation rates for ABP 215 and bevacizumab for all evaluated product-related variants and potency at 25°C, 40°C, and 50°C conditions were found to be similar. Figure 5.Size Exclusion High Performance Liquid Chromatography Profiles of ABP 215 and Bevacizumab Incubated at 50C for 14 days.
Biological and functional activity {#s0002-0006}
----------------------------------
The biological and functional activities of ABP 215 and bevacizumab evaluated in the similarity assessment are listed in [Table 1](#t0001){ref-type="table"}. Both Fab-mediated binding and functional activities, as well as Fc-mediated binding activities, were assessed. Since ABP 215 and bevacizumab do not exhibit effector functions,[@cit0011] the Fc-mediated binding activities were included to characterize the higher order structure of the molecules. The results from assays reporting quantitative data from the biological and functional assessment are summarized in [Table 3](#t0003){ref-type="table"}.
The biological functions that contribute to the clinical efficacy of bevacizumab are mediated by the neutralization of VEGF-A through the Fab domain. The similarity assessment included potency as measured by inhibition of proliferation in human umbilical vein endothelial cells (HUVEC) and relative binding to the target (VEGF-A). The results demonstrate that ABP 215 has similar potency and VEGF-A binding activity compared with bevacizumab. In addition, side-by-side comparison of binding affinity for VEGF-A further supports that ABP 215 has similar VEGF-A binding activity compared to bevacizumab. Furthermore, the blockade of signaling downstream of VEGF receptor 2 (VEGFR-2), the primary functional receptor for VEGF-A, was evaluated using a receptor tyrosine kinase (RTK) assay in HUVEC. The results support that ABP 215 demonstrates similar inhibition of VEGFR-2 signaling compared with bevacizumab ([Figure 6](#f0006){ref-type="fig"}). Figure 6.Inhibition of VEGFR-2 RTK autophosphorylation by ABP 215, bevacizumab (US), and bevacizumab (EU).
VEGF-A acts as a soluble ligand to induce angiogenesis, and, while bevacizumab binds VEGF-A, the antibody does not exhibit effector functions.[@cit0011] Binding assays were conducted to characterize the higher order structure of the two antibodies and particularly the Fc domains of ABP 215 and bevacizumab. The neonatal Fc receptor (FcRn) binds to IgG in the Fc region and mediates IgG homeostasis in humans. The results demonstrate that ABP 215 has similar FcRn binding compared with bevacizumab ([Table 3](#t0003){ref-type="table"}). Relative binding to multiple additional Fc gamma receptors with no established role in the efficacy or safety of ABP 215 and bevacizumab (FcγRIa, FcγRIIa, FcγRIIb, FcγRIIIa (158V), FcγRIIIa (158F) and FcγRIIIb) and the first subcomponent of the C1 complex of the classical pathway of complement activation (C1q) were also performed for characterization purposes. The results shown in [Table 3](#t0003){ref-type="table"} support that ABP 215 has similar relative Fc receptors and C1q binding compared to bevacizumab, thus confirming the minor differences observed in glycan levels did not affect binding activities in the Fc region. Overall, the biological activity results support the conclusion that ABP 215 is highly similar to bevacizumab.
General properties {#s0002-0007}
------------------
Following regulatory agency guidance, biosimilars must demonstrate that the "strength" is the same as the reference product.[@cit0016] The analytical similarity assessment of ABP 215 and bevacizumab for product strength was measured by protein concentration and volume and were found to be similar ([Table 2](#t0002){ref-type="table"}). In addition to the final dosage form strengths, other general properties, including osmolality, pH, appearance, color, and clarity, were also compared, and all results met the pre-specified criteria for analytical similarity.
Host-cell impurities {#s0002-0008}
--------------------
Any significant differences of the process-related impurities may have an adverse impact upon safety. The residual host cell proteins (HCP) in ABP 215 and bevacizumab were quantified and characterized by orthogonal methods. The HCP ELISA results using an assay specific to ABP 215 cell line and manufacturing process showed ABP 215 had similarly low levels of residual HCP compared to bevacizumab ([Table 2](#t0002){ref-type="table"}). To further characterize the type of HCPs present in samples, a reverse-phase two-dimensional liquid chromatography (2D-LC) coupled with online mass spectrometry involving data-independent acquisition (2D-LC-MS) method was used to identify and quantify HCPs by mass spectrometry.[@cit0017]^,^[@cit0018] Results show no new HCP was present in ABP 215 compared to bevacizumab. Finally, a 2D differential in-gel electrophoresis method comparing samples was performed to confirm the absence of new species in a qualitative manner. In addition, residual protein A and DNA were also in very low levels or below detection in both ABP 215 and bevacizumab.
Discussion {#s0003}
==========
ABP 215 was developed by Amgen Inc. and has been approved by the FDA and EMA as a biosimilar to bevacizumab. A comprehensive analytical strategy was developed and executed to assess the structural and functional similarity of ABP 215 and bevacizumab. The similarity assessment assessed critical and relevant attributes, including the primary structure, higher order structure, particles and aggregates, product-related substances and impurities, thermal forced degradation, biological activities, general properties, and process-related impurities for both ABP 215 and bevacizumab. The totality of the similarity results demonstrates that ABP 215 is analytically highly similar to bevacizumab. Although some minor differences in physicochemical attributes were observed between ABP 215 and bevacizumab, the biological and functional similarity results show these minor differences do not affect functions relevant to the mechanism of action of ABP 215 and bevacizumab. The results also demonstrate bevacizumab from both US and EU are analytically comparable. Thee analytical similarity assessment results provide the foundation for the scientific justification of extrapolation of the indications approved for bevacizumab to ABP 215.
Materials and methods {#s0004}
=====================
ABP 215 was manufactured by Amgen Inc. US-licensed bevacizumab manufactured by Genentech, Inc. (a Roche company), and EU-authorized bevacizumab manufactured by Roche were procured over a period of approximately 6 years. The reference products were stored and handled according to the manufacturer\'s instructions and tested as part of the analytical similarity assessment plan. ABP 215 was developed to reflect the same strength and presentations approved in the US and the EU.
This article does not contain any studies with human participants or animals. This paper was prepared according to the ICMJE Uniform Requirements and the International Society for Medical Publication Professionals\' "Good Publication Practice for Communicating Company-Sponsored Medical Research: The GPP2 Guidelines."
Intact and reduced deglycosylated mass analysis {#s0004-0001}
-----------------------------------------------
The molecular masses of intact molecules were determined by electrospray ionization-time of flight-mass spectrometer (ESI-TOF-MS) analysis. Samples were separated from buffer components and introduced to the mass spectrometer by size exclusion chromatography using an Ethylene Bridged Hybrid column. The resulting summed ion spectra were deconvoluted to produce molecular mass profiles. The theoretical mass calculations assumed no C-terminal lysine residues on the HC and that all the disulfides were intact.
The molecular masses of reduced and deglycosylated HC and LC were evaluated using ESI-TOF-MS to provide further assurance the polypeptide compositions were as expected. Samples were treated with PNGase F (Wako Chemicals) to remove N-linked glycans, subsequently denatured with guanidine hydrochloride (Thermo Scientific), and the disulfides reduced with tris (2-carboxyethyl) phosphine hydrochloride (Thermo Scientific). The samples were separated from buffer components and introduced to the mass spectrometer via size exclusion chromatography. Summed ion spectra were deconvoluted to produce molecular mass profiles, which were compared with theoretical mass values.
Reduced and non-reduced peptide map {#s0004-0002}
-----------------------------------
Reduced peptide map analysis was conducted by enzymatic digestion with trypsin (Roche Life Science). The sample was treated with dithiothreitol for reduction, sodium iodoacetic acid (IAA, Sigma-Aldrich) for alkylation, and PNGase F for the N linked glycan removal prior to digestion. The resulted peptides were separated by reversed phase UPLC using an increasing gradient of acetonitrile in water. The eluted peaks were detected by UV absorbance (214 nm) and the peptides were identified by on-line mass spectrometry (LC-MS/MS) using hybrid Ion Trap-Orbitrap Mass Spectrometer.
For non-reduced peptide map, samples were digested with trypsin under denaturing, but non-reducing conditions. The resulting peptides were analyzed by RP-HPLC using an increasing gradient of acetonitrile in water with UV detection at 214 nm. The peptides containing disulfide bonds were identified by comparing peptides generated under reducing and non-reducing conditions. Confirmation of peptide identity was achieved using an HPLC coupled with an electrospray mass spectrometer, allowing accurate determination of the molecular mass of each peptide.
Glycan map {#s0004-0003}
----------
N-linked glycans were evaluated by glycan map analysis using hydrophilic interaction liquid chromatography (HILIC) with fluorescence detection. Glycan mapping involved release of N-glycans through treatment with the amidase PNGase F. The reducing termini of the released glycans were then labeled through reductive amination with a fluorescent tag (2 aminobenzoic acid, 2-AA), and the labeled glycans were separated by HILIC. Peak identification was performed using mass spectrometry by coupling the HILIC HPLC with an ion-trap mass spectrometer for verification against the expected glycan mass.
Isoelectric points {#s0004-0004}
------------------
The pIs were determined by cIEF analysis. cIEF was performed on a capillary electrophoresis separation instrument equipped with a neutral-coated capillary (Beckman Coulter). Samples were electrophoresed through a pH gradient produced by an ampholytic solution (GE Healthcare) until they reached the pH equal to their pI and were then mobilized and detected by UV absorbance (280 nm) as they passed through a detection window in the capillary. The pIs of the sample peaks were determined using a linear regression between two pI marker peaks (Protein Simple).
Higher order structure by fourier-transform infrared spectroscopy and near ultraviolet-circular dichroism spectroscopy {#s0004-0005}
----------------------------------------------------------------------------------------------------------------------
Secondary structure was assessed by FTIR spectroscopy. The spectrum of the formulation buffer blank was recorded under identical conditions and subtracted from the protein solution spectra. The second derivative spectrum was calculated using a 9-point smoothing of the resulting spectra. Spectral similarity of testing molecule against reference molecule was calculated using the Thermo OMNIC software QC Compare function where 100% similarity indicates identical spectra.
Tertiary structure was assessed by near UV-CD spectroscopy. Protein solutions were diluted to approximately 0.7 mg/mL in test buffer for the near UV-CD measurements using cuvettes with a path length of 1 cm. The spectra were corrected for protein concentration and contributions from the buffer and were reported as CD ellipticity. Spectral similarity was calculated using the Thermo OMNIC software QC Compare function where 100% similarity indicates identical spectra.
Thermal stability by differential scanning calorimetry {#s0004-0006}
------------------------------------------------------
Thermal stability was assessed by DSC using a system in which temperature differences between the testing sample and buffer cells were continuously measured and calibrated. The unfolding of the protein molecules appears as an endothermic transition on the DSC thermogram, and is characterized by the thermal melting temperatures (T~m~). The protein concentration of the testing solutions was approximately 0.5 mg/mL, which was obtained by diluting the solutions in formulation buffer.
Subvisible particles by light obscuration and micro-flow imaging {#s0004-0007}
----------------------------------------------------------------
Subvisible particles were assessed by light obscuration (LO) using a HIAC 9703+ liquid particle counting system equipped with an HRLD 150 sensor. Particle concentration results were reported as cumulative particle counts per container for ≥ 2, ≥ 5, ≥ 10, and ≥ 25 µm size ranges.
Subvisible particles were also assessed by MFI particle imaging system containing a flow cell and a digital camera. Cumulative particle counts per mL for ≥ 5 μm particles were reported. To quantify product-related particles that are likely proteinaceous and thus have a higher risk for immunogenicity, the MFI data were further analyzed for the concentration of ≥ 5 μm non-spherical particles with an aspect ratio of \< 0.85.
Submicron particles by dynamic light scattering and field flow fractionation {#s0004-0008}
----------------------------------------------------------------------------
Submicron particles were assessed by dynamic light scattering. Protein solutions were diluted with formulation buffer to approximately 1 mg/mL, and triplicate measurements were taken for each test solution.
Submicron particles were also assessed by FFF. Product solutions were injected neat into the FFF system and UV and light scattering (LS) detectors were used to detect the presence of monomer, aggregates, and submicron particles. Polystyrene particle standards of 100 nm, 200 nm, and 500 nm were used as positive controls, and formulation buffer was used as the negative control.
Aggregates by analytical ultracentrifugation sedimentation velocity {#s0004-0009}
-------------------------------------------------------------------
Aggregates were assessed by AUC-SV. Product solutions were diluted with formulation buffer to approximately 0.5 mg/mL, and the AUC-SV measurements were taken at 45, 000 rpm, with absorbance at 280 nm as detection. Triplicate measurements were taken for each test solution.
Aggregates by size exclusion-high-performance liquid chromatography with light scattering detection {#s0004-0010}
---------------------------------------------------------------------------------------------------
Aggregates were also assessed by SE-HPLC-LS. The SE-HPLC method used employed an LS detector, a refractive index (RI) detector, and a UV detector at 280 nm. Product solutions were injected neat into the system at a load of approximately 280 µg. For molar mass calculation, an RI increment value of 0.185 mL/g was used. Results were reported as the molar mass of monomer, dimer, and high molecular weight species.
Size variants by size exclusion-high-performance chromatography, reduced and non-reduced capillary electrophoresis-sodium dodecyl sulfate {#s0004-0011}
-----------------------------------------------------------------------------------------------------------------------------------------
Native, or non-denatured, size variants were analyzed by size exclusion-HPLC (SE-HPLC). SE-HPLC measurements were made on an Agilent 1100 HPLC system with a Tosoh Bioscience TSK-GEL G3000SW column. Analytes were monitored by UV absorbance at 280 nm, and purity was evaluated by determining the peak area of each species as a percentage of the total peak area.
Capillary electrophoresis-sodium dodecyl sulfate (CE-SDS) was used for separation of denatured protein size variants under reduced or non-reduced conditions. For non-reduced condition, drug product samples were denatured using sodium dodecyl sulfate at 60^o^C for 5 minutes. For reduced condition, β-mercaptoethanol was added to the protein denaturation step to reduce the disulfide bonds and incubation was performed at 70^o^C for 10 minutes. After denaturation, both reduced and non-reduced samples were injected onto a bare, fused silica capillary and separated based on hydrodynamic size resulting from an applied electric field in which migration time of smaller size proteins is inversely related to overall size. Analytes were monitored by UV absorbance at 220 nm, and purity was evaluated by determining the peak area of each species as a percentage of the total peak area.
Charge variants by cation exchange-high-performance liquid chromatography {#s0004-0012}
-------------------------------------------------------------------------
Charged isoforms in drug product samples were separated on a CEX-HPLC Pro Pac WCX-10 analytical column. Eluted fractions using a salt gradient were monitored by UV absorbance at 280 nm and purity was evaluated by determining the peak area of each charged isoform group (acidic, main, and basic peaks) that eluted separately as a percentage of the total peak area.
Stability and degradation at accelerated conditions {#s0004-0013}
---------------------------------------------------
The thermal stability and degradation profiles of ABP 215 drug products and bevacizumab were determined at 25, 40, and 50^o^C. Subsequent characterization of the degradation samples was conducted using SE-HPLC, rCE-SDS, nrCE-SDS, CEX-HPLC, and potency assays.
Protein concentration {#s0004-0014}
---------------------
The protein concentration in the solution was determined by UV absorbance using the product extinction coefficient and the density values for both sample and formulation buffer.
Volume {#s0004-0015}
------
The analytical procedure for volume determination complies with USP \<1\>, PhEur 2.9.17, and JP 6.05. It is a quantitative method utilizing the product density and measuring its mass to calculate its volume.
Host cell impurities {#s0004-0016}
--------------------
The residual HCP ELISA method was developed specifically for the ABP 215 cell line and process. This method provides quantitative measurement of total HCP detected through the use of a proprietary anti-HCP polyclonal antibody reagent (Amgen).
Potency by HUVEC proliferation inhibition assay {#s0004-0017}
-----------------------------------------------
The proliferation inhibition bioassay is a quantitative cell-based assay utilizing HUVEC (Lonza) that measures the dose-dependent inhibitory effects of ABP 215 or bevacizumab on proliferation of cells expressing VEGFR. HUVEC express VEGFR-1, VEGFR-2 and VEGFR-3 receptors, as well as the co-receptors, neuropilin (NRP)-1 and NRP-2, which, upon interaction with VEGF-A, results in endothelial cell proliferation. In this method, HUVEC were incubated with varying concentrations of ABP 215 reference standard, control, and test samples in the presence of a constant concentration of VEGF-A (R&D Systems, Cat. No. 298-VS). After a timed incubation, an adenosine triphosphate (ATP) specific luminescent reagent (Promega) was added to the assay plates. Addition of this reagent resulted in cell lysis and generation of luminescence signal, which was proportional to the amount of ATP present. The amount of ATP present was directly proportional to the number of viable cells in the culture and inversely proportional to the concentration of tested samples. After assessing parallelism of the dose-response curves, the sample binding relative to the reference standard was determined using a 4-parameter logistic model fit using SoftMax® Pro Software (Molecular Devices). Results were reported as percent relative potency values.
VEGF-A binding by ELISA and SPR {#s0004-0018}
-------------------------------
A solid phase ELISA was used to determine binding to recombinant human VEGF-A. Recombinant VEGF-A (R&D Systems, Cat. No. 293-VE) was coated onto the wells of microtiter ELISA plates. A serial dilution of reference standard, control, and test sample(s) were added and incubated. Following a wash step, a goat anti-human IgG (Fc fragment) conjugated to horseradish peroxidase (HRP; ThermoFisher, Cat. No. 31413) was added to detect bound samples. After a final wash, a substrate/chromogen solution was added to the wells. The substrate changes color in the presence of HRP in proportion to the amount of ABP 215 or bevacizumab bound to VEGF-A. The reaction was stopped with 1.0 M phosphoric acid (Fisher Scientific) and absorbance was measured with a microplate reader. After assessing parallelism of the dose-response curves, the sample binding relative to the reference standard was determined using a 4-parameter logistic model fit using SoftMax® Pro Software (Molecular Devices). Results were reported as percent relative binding values.
The surface plasmon resonance (SPR) analysis was conducted at 25°C using a ProteOn XPR36 optical biosensor (Bio-Rad) equipped with a GLC sensor chip (Bio-Rad). Samples were captured on the GLC chip surface by a goat anti-human IgG1 capture antibody (Jackson ImmunoResearch Laboratories, Cat. No. 109-005-098) and immobilized using standard amine coupling chemistry. Recombinant human VEGF-A (R&D Systems, Cat. No. 298-VS) was injected at concentrations ranging from 50.0 to 3.13 nM and analyzed in triplicate. The data were aligned and double referenced using the ProteOn Manager 3.1.0 version 3.1.06 software (Bio-Rad). The data were then fit using Scrubber v2.0^©^ software (BioLogic), which is an SPR non-linear least squares regression fitting program. Results are reported as the average of three intra-assay replicates per lot. The association and dissociation phases for all VEGF-A concentrations were monitored for 240 seconds each. In addition, a long dissociation phase experiment of 5400 seconds was performed using the 50.0 nM VEGF-A concentration in order to better assess the slow dissociation rate of the antibodies. The binding kinetics were fit using a 1:1 binding model.
FcRn binding by AlphaScreen {#s0004-0019}
---------------------------
A two-step AlphaScreen® (Perkin Elmer) receptor binding assay was used to quantify the binding of human IgG Fc domain and FcRn. The assay measured the dose-dependent signal decrease observed when an Fc-containing sample is added to a reaction-containing FcRn-His (Amgen) and Fc-biotin (Amgen). Activity was determined by comparing the sample response to the response obtained for the reference standard. The sample binding relative to the reference standard was determined using a 4-parameter logistic model fit using SoftMax® Pro Software (Molecular Devices). Results were reported as percent relative binding values.
FcγRIa and FcγRIIIa by AlphaLISA {#s0004-0020}
--------------------------------
Relative binding to FcγRIa, FcγRIIIa (158V) and FcγRIIIa (158F) was quantified using AlphaLISA® assays (Perkin Elmer). FcγRIa, FcγRIIIa (158V) and FcγRIIIa (158F) GST-fusion proteins were generated at Amgen, Inc. Thousand Oaks, CA, USA. The AlphaLISA® assay is an Amplified Luminescent Proximity Homogenous Assay (Alpha) designed to measure the level of FcγR binding to the Fc portion of IgG1 antibodies. The assay contained two bead types, an acceptor bead and a donor bead. The acceptor bead binds recombinant human FcγR-glutathione-s-transferase (FcγR-GST). The donor beads were coated with a hydrogel that contains phthalocyanine, a photosensitizer, and streptavidin, which binds to biotinylated human IgG1 (Amgen). When bevacizumab or ABP 215 IgG1 is present at sufficient concentrations to inhibit the binding of FcγR-GST to the biotinylated human IgG1, a dose-dependent decrease in emission is observed using a plate reader. The sample binding relative to the reference standard was determined using a 4-parameter logistic model fit using SoftMax® Pro Software (Molecular Devices). Results were reported as percent relative binding values for either FcγRIa, FcγRIIIa (158V) and FcγRIIIa (158F).
FcγRIIa, FcγRIIb, and FcγRIIIb Binding by SPR {#s0004-0021}
---------------------------------------------
FcγRIIa, FcγRIIb, or FcγRIIIb (R&D Systems) was immobilized through standard amine coupling chemistry to a BiaCore CM5 chip (GE Healthcare Bio-Sciences). Binding data was collected 4 seconds prior to the end of injection for each concentration of a duplicate sample dilution series. The K~D~ was determined using the steady state affinity model. The binding data generated from the assay run was used to determine the relative binding activity of the test samples in comparison to the reference standard by dividing the K~D~ of the sample by that of the reference standard. Binding is reported as relative binding for FcγRIIa, FcγRIIb, and FcγRIIIb,
C1q binding by ELISA {#s0004-0022}
--------------------
A direct binding ELISA method was developed to assess the binding of ABP 215 to C1q. In this assay, bevacizumab or ABP 215 was adsorbed to a microtiter plate and incubated with C1q (Sigma-Aldrich, Cat. No. C1740). Bound C1q was detected with an anti-C1q-HRP conjugated antibody (Bio-Rad Hercules, Cat. No. 2221--5004P). After a final wash, a substrate/chromogen solution was added to the wells. The substrate changes color in the presence of HRP in proportion to the amount of C1q bound to ABP 215 or bevacizumab. The reaction was stopped with sulfuric acid (Sigma-Aldrich) and absorbance was measured with a microplate reader. The sample binding relative to the reference standard was determined using a 4-parameter logistic model fit using SoftMax® Pro Software (Molecular Devices). Results were reported as percent relative binding values for C1q.
Inhibition of VEGFR2 RTK autophosphorylation {#s0004-0023}
--------------------------------------------
HUVEC were incubated with varying concentrations of samples in the presence of a constant concentration of VEGF-A (R&D Systems, Cat. No. 298-VS-025/CF). After a timed incubation, the cells were lysed. VEGFR-2 was captured from the lysate onto streptavidin-coated Mesoscale Discovery (MSD, Cat. No. L15SA) plates using a biotinylated antibody against the extracellular portion of VEGFR-2 (R&D Systems, Cat. No. 05--321). This was followed by the addition of a murine anti-phosphotyrosine monoclonal antibody (EMD Millipore, Cat. No. 05--321) for detection of tyrosine phosphorylation on the captured VEGFR-2 and an anti-murine IgG conjugated with ruthenium (MSD, Cat. No. R32AC-1) for signal generation. The addition of a tripropylamine-containing buffer (MSD, Cat. No. R92TC-2) followed by electrical activation of the plate resulted in an electrochemiluminescent (ECL) signal detected by a plate reader. The ECL signal counts were proportional to the level of VEGFR-2 tyrosine phosphorylation.
Acknowledgments
===============
The authors would like to acknowledge the technical contributions of Ching Chen, Jill Crouse-Zeineddini, Nancy Jiao, Jennifer Rattan, Megan Verardo, Yunjen Young, and Qingchun Zhang,. Medical writing and editorial support was provided by Pasquale Iannuzzelli, PhD (Innovation Communications Group) under the direction of Monica Ramchandani, PhD (Amgen) and funded by Amgen, Inc.
Disclosure of potential conflicts of interest
=============================================
Neungseon Seo, Alla Polozova, Mingxuan Zhang, Zachary Yates, Shawn Cao, Huimin Li, Scott Kuhns, Gwendolyn Maher, Helen J. McBride, and Jennifer Liu are employees and stockholders of Amgen, Inc.
[^1]: Abbreviations: 2D-DIGE = 2-dimensional differential in-gel electrophoresis, 2D LC-MS = 2-dimensional liquid chromatography coupled with online mass spectrometry involving data-independent MS acquisition, ADCC = antibody-dependent cell-mediated cytotoxicity, AUC-SV = analytical ultracentrifugation sedimentation velocity, CDC = complement-dependent cytotoxicity, CEX-HPLC = cation exchange high performance liquid chromatography, cIEF = capillary isoelectric focusing, C1q = the first subcomponent of the C1 complex of the classical pathway of complement activation, DSC = differential scanning calorimetry, ELISA = enzyme-linked immunosorbent assay, FcR = fragment crystallizable receptor, FcγRIa = Fc gamma receptor Type Ia, FcγRIIa = Fc gamma receptor Type IIa, FcγRIIb = Fc gamma receptor Type IIb, FcγRIIIa = Fc gamma receptor Type IIIa, FcγRIIIb = Fc gamma receptor Type IIIb, FcRn = neonatal Fc receptor, FFF = field flow fractionation, FTIR = Fourier-transform infrared spectroscopy, HC = heavy chain, HCP = host cell protein, HILIC = hydrophilic interaction liquid chromatography, HPLC = high performance liquid chromatography, LC = light chain, MFI = micro-flow imaging, nrCE-SDS = non-reduced capillary electrophoresis -- sodium dodecyl sulfate, qPCR = quantitative polymerase chain reaction, rCE-SDS = reduced capillary electrophoresis -- sodium dodecyl sulfate, SE-HPLC-LS = size exclusion high performance liquid chromatography with light scattering, SE-HPLC = size exclusion high performance liquid chromatography, UV-CD = ultraviolet circular dichroism, VEGF-A = vascular endothelial growth factor A.
[^2]: n indicates a number of samples tested
[^3]: n indicates a number of samples tested
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Duck plague virus (DPV), or duck enteritis virus (DEV), is an important pathogen of ducks, which has caused serious losses in commercial duck production in domestic and wild waterfowl as a result of mortality, condemnations, and decreased egg production\[[@B1]\]. DPV is classified as the subfamily alphaherpesvirinae of the family herpesviridae based on the report of the Eighth International Committee on Taxonomy of Viruses (ICTV), but has not been grouped into any genus\[[@B2]\].
The genome of DPV is composed of a linear, double-stranded DNA, with 64.3% G+C content which is higher than any other reported avian herpesvirus in the subfamily alphaherpesvirinae\[[@B3]\]. To date, more and more DPV genes have been identified, such as UL24\[[@B4]-[@B6]\], UL31\[[@B7],[@B8]\], UL35\[[@B9],[@B10]\], UL51\[[@B11],[@B12]\], dUTPase\[[@B13]\], and gE\[[@B14]\] gene. However, the key genes and their functions remain to be elucidated, especially the viral envelope protein genes. Viral envelope proteins are particularly important because of their role in the virus-host relationship, including recognition, attachment and penetration of the virus into susceptible cells. In 2006, the DPV genomic library was successfully constructed in our laboratory, and one envelope protein gene, gC (NCBI GenBank accession no. [EU076811](EU076811)) was characterized\[[@B15]-[@B18]\], but the basic properties and biological functions of this envelope protein are not known.
gC is a major component of the virion envelope and is proved to be a multifunctional protein. gC homologues of herpes simplex virus type 1 (HSV-1), pseudorabies virus (PRV), and bovine herpesvirus type 1 (BHV-1), is the primary attachment protein, interacting with cell surface heparan sulfate proteoglycans (HSPG), thus mediating efficient virus attachment to the cells\[[@B19]-[@B24]\]. gC of HSV-1, varicella-zoster virus (VZV) and PRV\[[@B25]-[@B28]\] is also a major determinant for virulence. In case of Marek\'s disease virus (MDV), gC is required for horizontal transmission, together with US2, UL13 in combination\[[@B29]\]. Furthermore, gC has been demonstrated to be a critical immune evasion molecule, and the two glycoproteins, gC and gE, have a synergistic effect on mediating immune evasion\[[@B30],[@B31]\]. And gC of HSV-1 and -2, BHV-1, PRV, and Equine herpes virus types 1 and 4 (EHV-1 and -4) has been reported to bind complement component C3\[[@B32]-[@B34]\], thus modulating complement activation. Although nonessential for virus infectivity of cultured cells, gC is a highly antigenic glycoprotein, of which the importance in eliciting immune responses has been well documented for many herpesviruses\[[@B35]-[@B39]\]. However, whether the product of DPV gC gene shares these functions remains to be determined.
To begin addressing questions regarding gC properties or functions, we cloned and expressed the gC gene from DPV in the prokaryotic expression system, raised antiserum that recognizes the gC protein and revealed its temporal transcription course and subcellular localization in DPV-infected DEF cells. This work might provide a foundation for further studies on the function of DPV gC.
Results
=======
Cloning, prokaryotic expression and antigenicity analysis of the recombinant protein
------------------------------------------------------------------------------------
DPV gC gene from the genomic DNA was amplified and cloned into a T/A cloning vector pMD18-T, generating a recombinant cloning plasmid pMD18-T/gC (Figure [1](#F1){ref-type="fig"}). The recombinant plasmid was confirmed by DNA sequencing, PCR and restriction digestion (Figure [2a](#F2){ref-type="fig"}). The gC gene fragment, which was obtained by digestion of pMD18-T/gC with EcoRI and XhoI, was ligated into the fusion expression vector pET32a(+) (Figure [3](#F3){ref-type="fig"}) and identified by PCR and restriction digestion (Figure [2b](#F2){ref-type="fig"}). After confirmation, a positive clone was submitted to DNA sequencing and the result confirmed that the gC gene was in frame with the N-terminal His6 tag within the pET32a(+) multiple cloning sites (data not shown). Then this recombinant plasmid, pET32a-gC, was transformed into Escherichia coli BL21 (DE3) which, following induction with IPTG, expressed large quantities of the pET32a-gC recombinant protein (Figure [4a](#F4){ref-type="fig"}), and this recombinant protein was purified by gel and electric elution (Figure [4b](#F4){ref-type="fig"}). In order to examine the reactivity and specificity of the recombinant fusion protein, Western blot analysis was carried out. As shown in Figure [4b](#F4){ref-type="fig"}, the anti-DPV serum specifically recognized a 65 kDa band, which corresponded to the theoretical molecular mass of pET32a-gC.
{#F1}
{#F2}
{#F3}
{#F4}
Transcriptional analysis of DPV gC gene
---------------------------------------
FQ-PCRs were used to detect the expression of the DPV gC gene during viral infection. Total RNA was isolated from mock-or DPV-infected cells at indicated times, and then cDNA was synthesized using reverse transcriptase. Aliquots of cDNA at each time point were used as template for real-time PCR reactions containing primers either for gC gene or for β-actin. Table [1](#T1){ref-type="table"} presents data from experiments where the target (gC) and reference (β-actin) were amplified in separate wells. The 2^-ΔΔCt^method was used to calculate relative changes in the gene expression determined from quantitative real-time PCR experiments. As shown in Figure [5](#F5){ref-type="fig"} the level of gC mRNA had been increasing since 4 hpi, peaked between 28 and 36 hpi, and then declined.
######
Sample spreadsheet of data analysis using the 2-ΔΔCt method
Time gC β-actin ΔΔCt **2**^**-ΔΔCt**^ Log
------ ------ --------- ------ ------------------ ------- ---------- ------
1 h 21.3 24.5 10.6 13.7 -0.1 1.071773 0.03
4 h 20.7 24.5 11.3 13.7 -1.4 2.639015 0.42
7 h 19.6 24.5 12.5 13.7 -3.7 12.99603 1.11
10 h 16.5 24.5 12.5 13.7 -6.8 111.4304 2.05
14 h 14.1 24.5 10.7 13.7 -7.4 168.897 2.23
20 h 14.6 24.5 12.4 13.7 -8.6 388.0234 2.59
28 h 11.3 24.5 17.9 13.7 -17.4 172950.5 5.24
36 h 11.9 24.5 18.5 13.7 -17.4 172950.5 5.24
54 h 14.4 24.5 19.8 13.7 -16.2 75281.1 4.88
72 h 18.9 24.5 24 13.7 -15.9 61147.25 4.79
{#F5}
Time course expression of DPV gC protein
----------------------------------------
In this experiment, DEF cells were mock infected or infected with DPV, and at 4, 16, 32, 48, 60, 72 hpi, cell suspensions were harvested and lysed in RIPA buffer. Equal amounts of cell lysates were resolved by SDS-PAGE, and proteins on the gel were electrophoretically transferred to PVDF membrane and subjected to Western blot analysis with rabbit anti-DPV gC serum. The result, shown in Figure [6](#F6){ref-type="fig"} revealed that a 45 kDa protein was readily detected as early as 4 hpi and seemed to be present at increased levels at 48 hpi.
{#F6}
Subcellular localization
------------------------
The subcellular localization of gC protein was examined by indirect immunofluorescence staining. At various times after infection, DEF cells infected with DPV were fixed with 4% paraformaldehyde, treated with 5% bovine serum albumin (BSA) to block nonspecific binding and reacted with the DPV gC antiserum. Specific fluorescence became detectable only in the cytoplasm of infected cells as early as 4 hpi. At later times of infection, the protein converged at the perinuclear region of the cytoplasm. And after 60 hr, fluorescence was gently dispersed (Figure [7](#F7){ref-type="fig"}).
{#F7}
Discussion
==========
In this study we expand our initial observations, in which we identified and characterized the DPV gC gene and found that it has homologues in some herpesviruses sequenced to date. Analysis of the predicted 432 amino acid DPV gC protein indicates that it has several features common to other herpesvirus glycoproteins. There is a hydrophobic sequence of about 22 amino acids at the amino terminus that corresponds to a signal peptide and a 23-amino-acid hydrophobic sequence near the carboxy terminus that is predicted to span the membrane of the virus envelope. Amino acid sequence comparison revealed that DPV gC gene displayed similarities of 29.9%, 29.5%, 31.7% to this gene from MeHV-1, MDV1, MDV2, respectively. Further detailed analysis showed that it possessed a Marek\'s disease glycoprotein A conserved structural domain between the residues 170 and 425, indicating that the DPV gC and its counterpart in MDV may have similar functions.
As a first step toward the study of the gC protein, rabbit polyclonal antiserum specific to this protein were raised using an Escherichia coli BL21-produced recombinant gC fusion protein as antigen. For this purpose the plasmid pET32a-gC was constructed. When expressed in Escherichia coli BL21, this plasmid expresses the gC gene fragment along with a His6 tag attached to the N terminus. High levels of the resulting 65 kDa fusion protein were expressed in Escherichia coli BL21 following induction by IPTG. The induced fusion protein was purified as described in Methods.
To examine the reactivity and specificity of DPV gC protein, Western blot experiments were performed. The result showed that the fusion pET32a-gC protein was recognized by the rabbit anti-DPV IgG, indicating that the protein had good immunogenicity. Then the fusion pET32a-gC protein was used as antigen to produce the rabbit polyclonal antiserum specific for gC. The fusion pET32a-gC protein was recognized with the pET32a-gC antiserum by Western blot and the antiserum specifically reacted with a protein of approximately 45 kDa protein in DPV-infected DEF cells. These results indicated that the antiserum had a high level of reactivity and specificity. Therefore, we used this polyclonal antiserum for further experiments to characterize the gC protein of DPV.
During a productive infection of cultured cells, genes of herpesvirus have been found to be expressed in a temporally regulated cascade, in which immediate-early (IE) genes are expressed first, followed by early (E) genes and finally by late (L) genes\[[@B40]\]. Late genes are subdivided into two categories as leaky-late (γ~1~) or strict-late (γ~2~). The γ~1~genes can be suboptimally expressed in the absence of viral DNA synthesis, whereas the γ~2~, have a strict requirement for viral DNA synthesis. gC gene of many herpesviruses has been identified as a γ~2~gene, which is highly dependent upon the IE protein ICP27 during viral infection\[[@B41]-[@B47]\]. In HeLa cells infected with HSV-1 (5 PFU/cell), the transcript for the γ~2~gC was present from 2 to 8 hpi and the relative increase in gC transcript was detected by 8-h RNA hybridization\[[@B48]\]. Levine M\[[@B49]\] reported that gC protein was expressed at high levels within a single HSV replication cycle of about 10 to 14 h. From our data, the level of DPV gC mRNA had been increasing since 4 hpi, with maximal amounts between 28 and 36 hpi and the maximum gC expression was achieved by 48 hpi. These results demonstrated that the expression of this gene occurred at the late stage of infection, which was to some extent consistent with the results from the previous observations. In this report, DPV gC mRNA and protein peak levels were detected much later compared to HSV-1 gC gene, probably because of the difference in the cell type and the dose of infection.
Currently, little is known about the subcellular localization of the herpesviruses gC. To assemble clues to the function of the gene product, we investigated the subcellular localization of DPV gC in infected cells by indirect immunofluorescence experiments. The presented results showed that cytoplasm fluorescence first appeared in DPV-infected cells at 4 hpi. At later times of infection, the specific fluorescence was localized predominantly intracellularly in a perinuclear region which probably corresponds to the rough endoplasmic reticulum and/or Golgi apparatus of the infected cells, in which viral glycoproteins were synthesized and/or modificated.
Conclusions
===========
In this work, we characterized the gC gene of DPV, including the prokaryotic expression, antibody preparation, gene temporal transcription/translation course and subcellular localization. We found that the expression of this gene appeared at the late stage of viral infection and the gC protein showed a pronounced cytoplasmic staining in infected cells. These properties of the gC protein provide a foundation for further functional analysis of this gene.
Methods
=======
Cells and virus
---------------
Duck embryo fibroblasts (DEF) were cultured at 37°C with 5% CO~2~in minimal essential medium (MEM) containing 10% fetal bovine serum (FBS) (Hyclone, Logan, Utah, USA), 100 U/ml penicillin, and 100 μg/ml streptomycin.
DPV CH virulent strain was obtained from the Avian Disease Research Center of Sichuan Agricultural University. For infection, DPV of 2.2 × 10^7^TCID50/ml was employed.
After DPV inoculation, the DEF were incubated in MEM containing 3% FBS. Usually, the maximum virus titers could be obtained 72 h postinfection (hpi) when the cytopathic effect (CPE) was over 75%.
PCR amplification and plasmid construction
------------------------------------------
A pair of primers (5\'-CGGAATTCCAAAACGCCGCACAGATGAC-3\' and 5\'-CCCTCGAGGTATTCAAATAATATTGTCTGC-3\') was designed and used to amplify DPV gC gene from the genomic DNA. The amplified PCR product was cloned into a T/A cloning vector pMD18-T (TaKaRa), generating a recombinant cloning plasmid pMD18-T/gC (Figure [1](#F1){ref-type="fig"}). After verified by PCR, restriction analysis and DNA sequencing (TaKaRa), the gC gene fragment, which was obtained by digestion of pMD18-T/gC with EcoRI and XhoI, was ligated into prokaryotic vector pET32a(+) (Novagen) (Figure [3](#F3){ref-type="fig"}), which was digested previously with the same restriction enzymes. The recombinant plasmid, named pET32a-gC, was confirmed by PCR, restriction enzyme digestion and DNA sequencing (TaKaRa).
Prokaryotic expression, protein purification and antibody preparation
---------------------------------------------------------------------
pET32a-gC was transformed into Escherichia coli BL21 (DE3) and the bacteria were induced for 4 h with 0.6 mM IPTG at 37°C to express the fusion protein. The fusion protein was purified from inclusion bodies by gel and electric elution. To test the antigenicity of the recombinant fusion protein, proteins separated by 12% SDS-PAGE were subsequently subjected to Western blot analysis with rabbit anti-DPV serum. The purified recombinant protein was then mixed with an equal volume of Freund\'s complete adjuvant (Sigma) and used to immunize rabbits by intradermal injection, followed by two additional intradermal inoculations with Freund\'s incomplete adjuvant once every 7 days and the last inoculation with the purified recombinant protein. After the fourth immunization, anti-DPV gC serum was collected. Then, the purified IgG polyclonal antibodies were obtained by purification using caprylic acid and ammonium sulfate precipitation and High-Q anion exchange chromatography.
FQ-PCR
------
Total RNA was isolated from mock-or DPV-infected cells at different times (1, 4, 7, 10, 14, 20, 28, 36, 54, 72 hpi) and was reverse transcribed at 37°C for 120 min using random primer according to the manufacturer\'s instructions. The real-time PCR assays were performed using an iCycler iQ™ real-time PCR detection system (Bio-Rad Lab., Hercules, CA, USA). The primers used for PCR amplification were as follows: forward primer 5\'-GAAGGACGGAATGGTGGAAG-3\' and reverse primer 5\'-AGCGGGTAACGAGATCTAATATTGA-3\', which amplify a 78-base pair (bp) fragment of DPV gC gene, and for the endogenous control gene β-actin, forward primer 5\'-CCGGGCATCGCTGACA-3\' and reverse primer 5\'-GGATTCATCATACTCCTGCTTGCT-3\'. The amplification was performed in a 20 μl reaction mixture containing 9 μl of POWER High-Capacity cDNA Reverse Transcription Kits SYBR Green PCR master mix (Applied Biosystems), 0.5 μl of each primer, 1 μl of cDNA template and 9 μl of sterile ultra pure water. Three replicates of each reaction were performed. The PCR condition consisted of one cycle of 1 min at 95°C followed by 40 two-step cycles of 30 sec at 94°C and 30 sec at 60°C. Homogeneity of products from each reaction was confirmed by melt curve analysis. Analysis of the real-time PCR data was carried out using the comparative ΔΔCt method\[[@B50]\]. The fold change in expression of gC gene relative to the endogenous control gene (β-actin) at various time points was calculated as Fold change = Log(2^-ΔΔCt^), where ΔΔCt = (Ct, ~Target~-Ct, ~Reference~)~Time\ x~- (Ct, ~Target~-Ct, ~Reference~)~Time\ 0~.
Western blot analysis
---------------------
DEF were either mock infected or infected with DPV of 2.2 × 10^7^TCID50/ml, harvested at various indicated times (4, 16, 32, 48, 60, 72 hpi), lysed on ice for 30 min with an equal volume of radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, and 1 mmol/l phenylmethylsulfonyl fluoride) and centrifugated at 13,000×rpm for 15 min at 4°C\[[@B13]\]. Then equivalent amounts of the cell lysates were electrophoresed on 12% SDS-PAGE followed by staining with Coomassie Brilliant Blue R-250 and simultaneously electrophoretically transferred to a polyvinylidene fluoride (PVDF) membrane (Bio-Rad Lab., Hercus, CA, USA) in a transfer buffer at 120 V for 90 min. For Western blot analysis, purified rabbit polyclonal antibodies IgG was used as the primary antibody at a dilution of 1:50, followed by horse peroxidase (HRP) conjugated goat anti-rabbit IgG at a dilution of 1:5000 (KPL Inc., Gaithersburg, Maryland, USA) as the secondary antibody.
Subcellular localization
------------------------
DEF cells, grown on coverslips in six-well plates, were mock infected or infected with DPV of 2.2 × 10^7^TCID50/ml and then fixed with 4% paraformaldehyde for 15 min at room temperature at different times (4, 12, 24, 36, 48, 60 hpi). After blocking in PBS containing 5% bovine serum albumin (BSA) at 37°C for 1 h, the cells were incubated with purified rabbit polyclonal antibodies IgG (1:100 dilution) specific for recombinant DPV gC at 4°C overnight, rinsed three times for 10 min each with PBS and incubated with fluorescein isothiocyanate (FITC)-conjugated goat anti-rabbit IgG (Sino-American Biotechnology Co., Shanghai, China) for 1 h at 37°C. 4,6-Diamidino-2-phenylindole (DAPI; Sigma) staining was used to visualize the cell nuclei. Fluorescent images were viewed and recorded with the Bio-Rad MRC 1024 imaging system.
Competing interests
===================
The authors declare that they have no competing interests.
Authors\' contributions
=======================
BL carried out most of the experiments and drafted the manuscript. CX participated in the previous studies and helped to draft the manuscript. AC and MW critically revised the experiment design and the manuscript. DZ, QL, RJ, FB, ZC, YZ, ZY and XC helped with the experiment. All authors read and approved the final manuscript.
Acknowledgements
================
The research was supported by grants from the Changjiang Scholars and Innovative Research Team in University (PCSIRT0848), the earmarked fund for Modern Agro-industry Technology Research System (nycytx-45-12).
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Obesity and osteoporosis are two important global health problems with an increasing prevalence and a high impact on both mortality and morbidity.[@b1-dmso-4-273]--[@b4-dmso-4-273] Interestingly, during recent decades, both diseases have become a major health threat worldwide.[@b2-dmso-4-273] Age and female gender increase the risk of developing both obesity and osteoporosis, which affect millions of women.[@b3-dmso-4-273],[@b5-dmso-4-273]--[@b7-dmso-4-273] Age-related changes in body composition, metabolic factors, and hormonal levels after menopause, accompanied by a decline in physical activity, may all provide mechanisms for the propensity to gain weight and, in particular, for the increase in fat mass often characterized by replacement of lean mass by adipose tissue.[@b3-dmso-4-273],[@b4-dmso-4-273]
Obesity is due to an imbalance in which energy intake exceeds energy expenditure over a prolonged period.[@b2-dmso-4-273] In healthy adults, body weight is tightly regulated despite day-to-day variations in food intake and energy expenditure. Several environmental, nutritional, and hormonal factors appear to influence body weight.[@b1-dmso-4-273]--[@b3-dmso-4-273] For instance, postmenopausal women often show increased body weight, likely due to a decrease in basal metabolism, alteration of hormonal levels, and reduced physical activity.[@b8-dmso-4-273] Moreover, obese postmenopausal women are often affected by hypertension, dyslipidemia, diabetes mellitus, and cardiovascular disease, and have an increased risk of developing some cancers.[@b3-dmso-4-273],[@b9-dmso-4-273],[@b10-dmso-4-273] Interestingly, these women have always been considered protected against osteoporosis.[@b5-dmso-4-273],[@b6-dmso-4-273],[@b11-dmso-4-273]
Osteoporosis is a metabolic bone disease characterized by excessive skeletal fragility (due to a reduction in both bone quantity and quality), leading to an increased risk of developing spontaneous and traumatic bone fractures.[@b7-dmso-4-273] More than 40% of postmeno-pausal women, on average, will suffer at least one osteoporosis fracture, often leading to permanent and severe disability, nursing home placement, and even death.[@b11-dmso-4-273],[@b12-dmso-4-273] The rate of bone loss in adults reflects the interaction between genetic and environmental factors, which also influences the extent of bone acquisition during growth, known as peak bone mass.[@b13-dmso-4-273]
It is known that fractures in childhood have been associated with alterations in body composition, such as increased adiposity and bone structure, suggesting that these might be the earliest signs of skeletal insufficiency.[@b14-dmso-4-273] Soon after menopause, the process of bone loss begins in women, due to increased bone resorption by osteoclasts, which overcomes bone formation by osteoblasts.[@b13-dmso-4-273] Moreover, osteoblast function declines with aging, determining the imbalance between bone resorption and bone formation.[@b15-dmso-4-273] Traditionally, osteoporosis has been regarded as a disorder associated only with fracture and skeletal disability in old age, but recent studies demonstrate that bone mineral density appears to be a better long-term predictor of death than blood pressure or cholesterol.[@b12-dmso-4-273],[@b16-dmso-4-273] Further data published in recent decades indicate that low bone mineral density is a strong and independent predictor of all-cause mortality, including cardiovascular mortality.[@b12-dmso-4-273],[@b16-dmso-4-273]
Body fat and lean mass are correlated with bone mineral density, with obesity apparently confering protection against bone loss after menopause.[@b5-dmso-4-273],[@b6-dmso-4-273],[@b17-dmso-4-273] The pathophysiological role of adipose tissue in skeletal homeostasis probably lies in the role that several adipokines play in bone remodeling via their effects on either bone formation or resorption. Since the demonstration that bone cells express several specific hormone receptors, the skeleton has come to be considered an endocrine target organ.[@b18-dmso-4-273]--[@b21-dmso-4-273] Additionally, recent observations have shown that bone-derived factors, such as osteocalcin and osteopontin, may affect body weight control and glucose homeostasis,[@b22-dmso-4-273]--[@b24-dmso-4-273] suggesting a possible role of bone tissue as an endocrine organ with the presence of a potential feedback mechanism between the skeleton and endocrine organs.[@b25-dmso-4-273] Thus, the cross-talk between fat and bone likely constitutes a homoeostatic feedback system in which adipokines and molecules secreted by osteoblasts and osteoclasts represent the link of an active bone-adipose axis. However, the mechanism(s) by which all these events occur remains unclear.
Fat and bone correlation: evidence-based observations
=====================================================
In the last three decades, the association between obesity and osteoporosis has been actively investigated from epidemiological, clinical, and basic research points of view, and common pathophysiological links have been proposed: both obesity and osteoporosis are influenced by genetic and environmental factors, or the interaction between them; aging is associated with both diseases and with a high incidence of bone loss and bone marrow adiposity; bone remodeling and adiposity are both regulated via a complex interplay of adipokines and hormones; and adipocytes and osteoblasts derive from a common progenitor, ie, the mesenchymal stem cell,[@b11-dmso-4-273] as shown in [Figure 1](#f1-dmso-4-273){ref-type="fig"}.
Extensive data have shown that, in healthy pre-menopausal and postmenopausal women, total body fat is positively related to bone mineral density, an important and measurable determinant of fracture risk,[@b26-dmso-4-273]--[@b27-dmso-4-273] that high body weight (or body mass index) is correlated with high bone mineral density, and that decreased body weight leads to bone loss.[@b28-dmso-4-273]--[@b32-dmso-4-273] Furthermore, fat mass, the most important index of obesity, has been demonstrated to have a similarly beneficial effect, leading to an increase in bone mass,[@b17-dmso-4-273],[@b33-dmso-4-273] while a beneficial effect of fat mass on bone mineral density is confirmed in white women but not in white men.[@b34-dmso-4-273]
Although these data indicate that obesity exerts a protective effect on bone tissue, more recent studies have described an opposite event. In particular, although cross-sectional and longitudinal studies have shown that bone mass is positively related to body weight and body mass index, there are controversial issues as to whether lean mass or fat mass might be the most important determinant of bone mineral density.[@b6-dmso-4-273] In particular, the evidence suggests an inverse relationship between obesity and osteoporosis depending on how obesity is defined. In the studies where obesity is defined on the basis of body mass index or body weight, obesity appears to act as a protective factor against bone loss and fractures; however, if obesity is considered as a percentage of body fat and distribution, as in the study published by Zhao et al in a Chinese population,[@b11-dmso-4-273] it becomes a risk factor for osteoporosis.
In particular, there are data indicating that women with a high body mass index (25--29.9 kg/m^2^) are protected from osteoporosis, but there is increasing evidence conflicting with this observation, suggesting that obesity (body mass index \> 30) might actually interfere with bone health.[@b11-dmso-4-273]
In accordance with the data reported by Zhao et al,[@b11-dmso-4-273] Hsu et al showed that matching of Chinese subjects by body mass index, across 5 kg strata of body weight, revealed a negative relationship between fat and bone mass, and the risk of osteoporosis and nonspinal fractures was significantly higher for subjects with a higher proportion of body fat, independent of body weight.[@b35-dmso-4-273] Our group has recently demonstrated that 37% of 395 obese adult subjects had significant skeletal changes. In particular, this subpopulation showed a lower bone mineral density at the lumbar spine than expected for both their young age and high body mass index.[@b36-dmso-4-273] Further characterization showed that different grades of adiposity could affect skeletal health status differently. In fact, stratification of the population into three different groups according to body mass index status, showed a slightly different bone mineral density pattern among the groups. Overweight subjects (body mass index 26--29) did not show any change in skeletal health, while obese and severely obese subjects (body mass index \> 30) had significant alteration in their bone mineral density levels, with an increased number of individuals having a lower bone mass than would be expected for their age and body weight.[@b36-dmso-4-273] Evaluation of hormonal, metabolic, and lipid profiles did not show significant differences between the groups, athough more detailed analysis seemed to show higher inflammatory markers and lower levels of circulating vitamin D (Migliaccio et al, unpublished data). Indeed, data published by Blum et al from a cohort of 153 premenopausal women demonstrated that a high amount of fat mass is negatively associated with bone mass.[@b37-dmso-4-273] Thus, all these data suggest an important role for fat distribution as total fat mass itself. A recent study of 907 healthy postmenopausal women by Kim et al demonstrated that body weight was positively related to bone mineral density and vertebral fracture risk, whereas percentage of body fat and waist circumference were related to a low bone mineral density and to a higher risk for vertebral fractures.[@b38-dmso-4-273]
Even racial differences appear to influence fat and bone interaction. Castro et al reported that obesity is negatively associated with bone mineral density in black women, but not in white women,[@b39-dmso-4-273] while Afghani and Goran reported an inverse correlation between subcutaneous abdominal adipose tissue and bone mineral density in whites, but not in blacks. In the same study, the authors reported an inverse association between visceral fat and bone mineral density in blacks, but not in whites.[@b40-dmso-4-273] These conflicting results suggest a complex effect of fat mass on bone tissue related to sample size, ethnicity, gender, study design, methods of statistical analysis, and population structure. Nevertheless, several lines of evidence from environmental and medical interventions support an inverse correlation between fat and bone mass, ie, physical exercise increases bone mass while reducing fat mass,[@b41-dmso-4-273] supplementation with calcium and vitamin D appears beneficial for the prevention of both osteoporosis and obesity,[@b42-dmso-4-273] and menopause is also associated with increased fat mass, increased bone loss, and decreased lean mass.[@b43-dmso-4-273] Estrogen replacement therapy in postmenopausal women improves both lean mass and bone mass, and reverses menopause-related weight gain.[@b44-dmso-4-273] Whereas estrogens reduce the risk of bone loss and obesity, other pharmacological interventions have been shown to increase both osteoporosis and obesity, such as treatment with gonadotropin-releasing hormone agonists and the use of glucocorticoids.[@b45-dmso-4-273]--[@b48-dmso-4-273] Additionally, recent findings have indicated that some antidiabetic drugs, which interfere with peroxisome proliferator-activated receptor gamma (PPARγ) and thus with adipocyte differentiation, also appear to influence skeletal homeostasis and fracture risk significantly.
Thiazolidinedione and other selective PPARγ agonists, such as rosiglitazone and pioglitazone, play a prominent role in the treatment of type 2 diabetic patients. In vitro analyses demonstrate that various PPARγ ligands not only induce murine bone marrow stromal cell adipogenesis, but also inhibit osteogenesis,[@b49-dmso-4-273] and in vivo studies demonstrate that PPARγ agonists reduce bone mineral density and increase fracture rates, notably distal extremity fractures in female type 2 diabetic patients.[@b50-dmso-4-273]
Fat and bone correlation: potential mechanisms of interaction
=============================================================
Several potential mechanisms have been proposed to explain the complex relationship between adipose tissue and bone tissue. Fat has long been viewed as a passive energy reservoir, but since the discovery of leptin and identification of other adipose tissue-derived hormones and serum mediators,[@b51-dmso-4-273]--[@b53-dmso-4-273] fat has come to be considered as an active endocrine organ which modulates energy homeostasis. Adipose tissue also secretes various inflammatory cytokines, including interleukin (IL)-6 and tumor necrosis factor-alpha,[@b54-dmso-4-273] and altered production of these proinflammatory mediators is thought to have adverse metabolic and cardiovascular consequences. All these molecules, which include resistin, leptin, adiponectin, and IL-6, affect human energy homeostasis and may well be involved in bone metabolism, contributing to the complex relationship between adipose tissue and bone tissue ([Figure 2](#f2-dmso-4-273){ref-type="fig"}).[@b55-dmso-4-273]
Fat tissue is one of the major sources of aromatase, an enzyme also expressed in the gonads, which synthesizes estrogens from androgen precursors. Estrogens are steroid hormones which play a pivotal role in the maintenance of skeletal homeostasis, protecting against osteoporosis by reducing bone resorption and stimulating bone formation. This extragonadal estrogen synthesis in fat tissue becomes the dominant estrogen source in postmenopausal women, due to the lack of ovarian function.[@b6-dmso-4-273] Additionally, in obese post-menopausal women, increased estrogen synthesis by adipose tissue has been suggested as one of the potential mechanisms for the protective effect of fat mass on bone.
On the other hand, studies in humans lacking aromatase and in estrogen receptor-α and receptor-β knockout mice indicate that estrogens protect against bone loss, and support the hypothesis that these hormones may inhibit the development of obesity,[@b56-dmso-4-273]--[@b61-dmso-4-273] as suggested also by the prevention of menopause-induced fat mass gain[@b62-dmso-4-273]--[@b64-dmso-4-273] and reduction of the incidence of osteoporotic fractures by estrogen replacement therapy.[@b65-dmso-4-273] In support of this hypothesis, decreased endogenous estrogen levels have been shown to be coupled with an increase in adipocyte numbers and decreased osteoblast counts in the bone marrow of postmenopausal women.[@b66-dmso-4-273]
As mentioned above, several adipokines are involved in the fat-bone interaction. Leptin suppresses appetite, increases energy expenditure, and regulates bone remodeling, and is the most important adipocyte-derived hormone.[@b67-dmso-4-273]--[@b70-dmso-4-273] The effect of leptin on bone is complex, and both negative[@b37-dmso-4-273],[@b71-dmso-4-273] and positive actions[@b72-dmso-4-273]--[@b74-dmso-4-273] on bone mineral density have been reported in humans. Leptin-deficient ob/ob mice and leptin receptor-deficient db/db mice are extremely obese, with increased vertebral trabecular bone volume due to increased bone formation, despite hypogonadism and hypercortisolism.[@b69-dmso-4-273] Interestingly, intracerebroventricular infusion of leptin in both ob/ob and wild-type mice was shown to decrease vertebral trabecular bone mass.[@b69-dmso-4-273] In vivo studies indicate that the effect of leptin may depend on its site and mode of action,[@b75-dmso-4-273]--[@b77-dmso-4-273] and it has been proposed that peripheral administration of leptin could increase bone mass by inhibiting bone resorption[@b78-dmso-4-273] and increasing bone formation,[@b52-dmso-4-273],[@b79-dmso-4-273] while inhibiting bone formation through a central nervous system effect.[@b69-dmso-4-273] In vitro studies also found that leptin can act directly on bone marrow-derived mesenchymal stem cells to enhance their differentiation into osteoblasts and to inhibit their differentiation into adipocytes.[@b79-dmso-4-273],[@b80-dmso-4-273]
Takeda et al expanded these observations further, demonstrating that the effects of intracerebroventricular leptin are mediated by the sympathetic nervous system, and that osteoblasts express β-adrenergic receptors via which it is probable that administration of β-adrenergic agonists decreases trabecular bone volume by inhibiting bone formation.[@b70-dmso-4-273] Noradrenaline also seems to increase bone resorption by promoting nuclear factor kappa-B ligand expression and inhibiting osteoblast proliferation.[@b81-dmso-4-273]--[@b83-dmso-4-273] The effect of leptin on obesity is mediated via proopiomelanocortin neurons and neuropeptide Y.[@b84-dmso-4-273] Leptin reduces food intake and increases energy expenditure by stimulating proopiomelanocortin neurons to secrete α-melanocyte-stimulating hormone,[@b85-dmso-4-273] which, in addition to having an effect on obesity, might contribute to bone resorption[@b82-dmso-4-273] but not bone formation.[@b70-dmso-4-273] Interestingly, type 4 α-melanocyte-stimulating hormone receptor knockout mice have high bone mass due to decreased bone resorption.[@b82-dmso-4-273]
Leptin also inhibits expression of neuropeptide Y, a hypothalamus-derived peptide, essential for the regulation of food consumption, energy homeostasis, and bone remodeling.[@b86-dmso-4-273],[@b87-dmso-4-273] Hypothalamus-specific *NPY* knockout mice show a significant decrease in body weight, a significant increase in food intake, and a two-fold increase in trabecular bone volume compared with wild-type animals.[@b88-dmso-4-273],[@b89-dmso-4-273]
Adiponectin is another adipocyte-derived hormone which has anti-inflammatory and antiatherogenic effects, regulating energy homeostasis and bone remodeling.[@b90-dmso-4-273]--[@b94-dmso-4-273] In contrast with leptin, serum adiponectin levels are reduced in obese and diabetic subjects[@b95-dmso-4-273] and increase after weight loss.[@b92-dmso-4-273] Human osteoblasts express adiponectin and its receptors,[@b92-dmso-4-273] but both negative and positive links between adiponectin and bone mineral density have been reported.[@b96-dmso-4-273],[@b97-dmso-4-273] Other in vivo and in vitro studies show that adiponectin increases bone mass by suppressing osteoclastogenesis and activating osteoblastogenesis,[@b92-dmso-4-273]--[@b94-dmso-4-273] suggesting that a rise in adiponectin levels caused by fat reduction could have a beneficial effect on bone mineral density.
Thommesen et al showed that resistin may play a role in bone remodeling, indicating that it is expressed in mesenchymal stem cells, osteoblasts, and osteoclasts in bone marrow. Resistin increases osteoblast proliferation and cytokine release, as well as osteoclast differentiation,[@b98-dmso-4-273] so the effect of resistin on bone is still unclear, and further studies are needed to understand its role better.
IL-6 is a pluripotent inflammatory cytokine, released from adipocytes, adipose tissue matrix, osteoblast, and elsewhere in the body.[@b99-dmso-4-273] In particular, adipose tissue accounts for one-third of circulating levels of IL-6. Obese subjects have high circulating levels of this proinflammatory cytokine,[@b100-dmso-4-273],[@b101-dmso-4-273] and genetic polymorphism of IL-6 is associated with obesity.[@b102-dmso-4-273] Moreover, peripheral administration of IL-6 induces hyperlipidemia, hyperglycemia, and insulin resistance in rodents and humans.[@b103-dmso-4-273] In contrast, administration of IL-6 in the central nervous system increases energy expenditure and decreases body fat in rodents.[@b103-dmso-4-273] IL-6 is also a well recognized stimulator of osteoclastogenesis and bone resorption[@b104-dmso-4-273],[@b105-dmso-4-273] but some data show that IL-6 mRNA is expressed in preosteoblasts and osteoblasts,[@b106-dmso-4-273] and that IL-6 stimulates osteoblast proliferation and differentiation[@b107-dmso-4-273] by controlling the production of local factors,[@b108-dmso-4-273] and it might play a role in bone formation in conditions of high bone turnover.[@b108-dmso-4-273],[@b109-dmso-4-273]
In addition to adipocytes, adipose tissue contains various stromal and vascular cells, including fibroblasts, vascular endothelial cells, and inflammatory cells. Adipocytes were initially thought to be the major source of adipose-derived mediators, but recent studies have shown that macrophages infiltrate adipose tissue, and that these macrophages, along with other cells that reside in the stroma, also contribute to the production and secretion of humoral mediators, particularly inflammatory cytokines.[@b54-dmso-4-273] A paracrine loop involving free fatty acids and inflammatory cytokines has been postulated to establish a vicious cycle between adipocytes and macrophages, thereby propagating inflammation.[@b110-dmso-4-273],[@b111-dmso-4-273] Therefore, it is important to define interactions between adipocytes, osteoblasts, and stromal cells in obese subjects.
Adipocytes and osteoblasts: a common origin
===========================================
Adipocytes and osteoblasts originate from a common progenitor, ie, a pluripotential mesenchymal stem cell,[@b112-dmso-4-273] which has an equal propensity for differentiation into adipocytes or osteoblasts (or other lines) under the influence of several cell-derived transcription factors. This process is complex, suggesting significant plasticity and multifaceted mechanism(s) of regulation within different cell lineages, among which are adipocytes and osteoblasts.[@b113-dmso-4-273],[@b114-dmso-4-273] Several studies have examined the function of adipocytes in bone marrow. Mesenchymal stem cells isolated from marrow in postmenopausal osteoporotic patients express more adipose differentiation markers than those from subjects with normal bone mass.[@b115-dmso-4-273]
As mentioned earlier, adipocytes secrete endocrine and paracrine factors that strongly influence bone differentiation and remodeling. Estrogens are among these factors, explaining why increased body weight in postmenopausal women is associated with slower rates of bone loss.[@b113-dmso-4-273],[@b116-dmso-4-273],[@b117-dmso-4-273] However, the interaction between estrogens and fat appears to be complex. Martin and Zissimos showed pronounced fatty infiltration in the bone marrow of rats following oophorectomy, suggesting a pivotal role of estrogen in regulating adipocyte recruitment.[@b118-dmso-4-273] On the other hand, the presence of aromatase in fat cells allows higher intramarrow conversion of testosterone into estrogens which, in turn, can inhibit bone resorption.[@b113-dmso-4-273]
The effect of estrogen on bone and adipose tissue formation has long been recognized in experimental animal models.[@b118-dmso-4-273],[@b119-dmso-4-273] In humans, changes in estrogen status due to advancing age and menopause have been correlated with increased levels of IL-6 and IL-11, which are both associated with bone loss.[@b120-dmso-4-273] It is interesting to speculate whether the increase in adipogenesis subsequent to menopause is due to a relief of repression or to an induction of the adipogenic phenotype, even though in vitro data suggest that the default "switch" might be adipogenesis, a process which might normally be inhibited in vivo prior to estrogen depletion.[@b24-dmso-4-273]
Other members of the nuclear hormone receptor family contribute to control of adipogenesis and osteogenesis. PPARγ plays a central role in initiating adipogenesis.[@b113-dmso-4-273] Mutations of the PPARγ gene are associated with an altered balance between bone and fat formation in the bone marrow. The nuclear hormone receptor family of transcriptional regulatory proteins is activated by a range of ligands, including steroid hormones, naturally occurring metabolites, synthetic chemicals, and as yet unidentified endogenous compounds (orphan receptors).
Thiazolidinedione and other PPARγ ligands, such as rosiglitazone and pioglitazone, play a prominent role in the treatment of type 2 diabetes. However, in vitro analyses demonstrate that various PPARγ ligands not only induce murine bone marrow stromal cell adipogenesis but also inhibit osteogenesis.[@b49-dmso-4-273] In particular, PPARγ-2 is the dominant regulator of adipogenesis, and ligand activation of PPARγ-2 favors differentiation of mesenchymal stem cells into adipocytes rather than into osteoblasts.[@b116-dmso-4-273] Akune et al showed that PPARγ insufficiency led to increased osteoblastogenesis in vitro and higher trabecular bone volume in vivo, confirming the key role of mesenchymal stem cell lineage allocation in the skeleton.[@b112-dmso-4-273] Interestingly, aged mice exhibit fat infiltration into bone marrow and enhanced expression of PPARγ-2, along with reduced mRNA expression of bone differentiation factors.[@b121-dmso-4-273] Mice with premature aging (the SAM-P/6 model) show nearly identical patterns of adipocyte infiltration, with impaired osteoblastogenesis,[@b122-dmso-4-273] indicating that aging, or events that accelerate aging, result in significant bone marrow adiposity and a defect in osteoblastogenesis in mice.[@b123-dmso-4-273]
The Wnt signaling pathway works in a coordinated manner with other transmembrane signals, including multiple ligands, antagonists, receptors, coreceptors, and transcriptional mediators, such as β-catenin.[@b124-dmso-4-273] Specific elements of the Wnt signaling pathway have been found to inhibit adipogenesis[@b125-dmso-4-273],[@b126-dmso-4-273] while promoting osteogenesis.[@b127-dmso-4-273]--[@b131-dmso-4-273] Wnt inhibition of adipogenesis is mediated via β-catenin, which interferes with PPARγ transcriptional activation of downstream targets.[@b132-dmso-4-273] Following exposure to transforming growth factor beta, human bone marrow mesenchymal stem cells increase their expression of various Wnt receptors and ligands.[@b133-dmso-4-273]
Members of the epidermal growth factor family, such as protein Pref-1, influence both adipogenesis and osteogenesis. In vitro analysis of human bone marrow mesenchymal stem cells has shown that Pref-1 overexpression blocks both adipogenesis and osteogenesis. This finding is consistent with the hypothesis that Pref-1 maintains mesenchymal stem cells in a multipotent state.[@b134-dmso-4-273]
Further experimental tools, such as gene microarrays, are being used to document the relationship between classical steroid hormones and bone and fat formation in the marrow. One study has examined the skeletal phenotype of mice deficient in both thyroid receptors α and β. These mice showed increased mRNA levels for adipocyte-specific genes, increased numbers of bone marrow adipocytes, and reduced trabecular and total bone mineral density.[@b135-dmso-4-273] The inbred SAM-P/6 murine strain provides a model of accelerated senescence characterized by osteopenia and increased fat mass in bone marrow.[@b136-dmso-4-273] Recent studies have found that 1,25(OH)2 vitamin D treatment inhibits adipogenesis and enhances osteogenesis in SAM-P/6 mice, with a 50% reduction in PPARγ mRNA and protein levels.[@b123-dmso-4-273] Moreover, gene microarray analyses demonstrated coordinated induction of osteoblastogenic genes and a reduction of adipogenic genes after 1,25 (OH)2 vitamin D treatment, which stimulates not only bone formation but also bone resorption, according to circulating biomarkers of bone turnover.[@b137-dmso-4-273] Overall, these recent findings involving classical steroid receptors support the inverse relationship between adipogenic and osteogenic differentiation in the bone marrow microenvironment. This is mediated, in part, by cross-talk between the pathways activated by steroid receptors, PPARs, and other cytokines and paracrine factors.
Finally, other factors, such as total caloric intake, type of nutrients, alcohol consumption, oxygen tension, and cellular oxidation-reduction pathways influence bone marrow adipogenesis despite osteoblastogenesis,[@b24-dmso-4-273] showing that the bone marrow mesenchymal stem cell may consider multiple differentiation pathways during its lifetime and, indeed, may dedifferentiate and transdifferentiate in response to changes in the microenvironment.
Conclusion
==========
Obesity and osteoporosis are two major global health problems with an increasing prevalence and a high impact on mortality and morbidity. Menopause is characterized by the cessation of ovarian estrogen production and is associated with increased bone loss, increased fat mass, and decreased lean mass. Age-related changes in hormone levels, in association with changes in body composition, metabolic factors, and decreased physical activity, probably provide the mechanisms for the propensity to postmenopausal gain of fat mass, and thus increased body weight. Estrogen synthesis in adipose tissue, through aromatase, becomes the dominant estrogen source in postmenopausal women. In particular, in obese postmenopausal women, increased estrogen synthesis by fat tissue has been suggested as one of the potential mechanisms for the protective effect of fat mass on bone.
Even though there are data indicating that women with high body mass index are protected from osteoporosis, increasing evidence seems to show conflicting results regarding this issue, suggesting that obesity might actually interfere with bone health. In particular, the relationship between obesity and osteoporosis depends on how obesity is defined. If obesity is defined on the basis of body mass index or body weight, it appears to protect against bone loss and fractures. However, if obesity is based on the percentage of body fat, it may be a risk factor for osteoporosis.
The existence of a cross-talk between fat and the skeleton suggests a homoeostatic feedback system in which adipokines and bone-derived molecules form part of an active bone-adipose axis. However, the mechanism(s) by which all these events occur remains unclear. Of course, further basic science research and epidemiological studies with large sample sizes, robust study design, and careful data analysis will be needed to show the true effect of fat mass on bone.
The relationship between fat mass and bone is confounded by complex genetic backgrounds and by interactions between metabolic factors and regulatory pathways influencing both obesity and osteoporosis. Although the evidence that adipose tissue exerts a protective effect against bone loss is greater than that showing a negative association, the recent increasing data on this issue suggest that both obese and nonobese postmenopausal women should be considered at risk for alteration in bone mineral density and osteoporosis. Specific and careful characterization of skeletal metabolism and further studies evaluating skeleton changes may be useful in obese women, because aging itself might also increase their risk of developing fractures later in life.
**Disclosure**
The authors report no conflicts of interest in this work.
{#f1-dmso-4-273}
{#f2-dmso-4-273}
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Cognitive impairment is a frequently reported issue for patients with fragile X syndrome (FXS), which is the most common form of inherited intellectual disability, and the leading single gene cause of autism spectrum disorder ([@B37]; [@B26]; [@B1]). This syndrome is caused by the transcriptional silencing of the fragile X mental retardation 1 (*Fmr1*) gene, resulting in the loss or mutation of its product, fragile X mental retardation protein (FMRP), a RNA-binding protein that associates with polyribosomes and regulates translation ([@B7]; [@B56]). FMRP has been shown to repress the translation of several targets, including proteins critical for synaptic function ([@B56]; [@B11]). Of particular interest for the current study is the identification of interaction partners that play roles in signaling pathways related to synaptic plasticity ([@B43]) and dendritic structure ([@B39]; [@B12]).
Using *Fmr1* knockout (*Fmr1* KO) mice, we and others have demonstrated that the cognitive impairments in FXS may be linked to a disruption in *N*-methyl-[D]{.smallcaps}-aspartate receptor (NMDAR)-dependent synaptic plasticity in the hippocampal dentate gyrus (DG) ([@B54]; [@B15]; [@B18]; [@B6], [@B5]). The NMDA receptor forms a heterotetramer between two obligatory GluN1 subunits and two GluN2 subunits. The GluN2 subunits are differentially expressed during development, with GluN2B subunits initially being more highly expressed than GluN2A subunits early in neuronal development ([@B50]). The impact of NMDAR on dendritic structure can also have functional implications, as neuronal models indicate that dendritic morphology can have a significant impact on neuronal firing patterns ([@B31]).
Activation of NMDARs is known to play an important role in dendritic arborization and spine morphogenesis during neuronal development ([@B44]). NMDARs appear to contribute to spine and dendrite formation; however their exact role remains controversial. Using specific GluN2 antagonists in cultured cells, GluN2A subunits have been associated with dendritic arborization, while GluN2B subunits were associated with dendritic spine formation ([@B25]). In contrast, introducing the GluN2B subunits into ventral spinal neurons in culture enhanced dendritic arborization, an effect not observed with GluN2A subunits ([@B40]). Recently we have shown that genetic deletion of the Glun2A subunit significantly decreases dendritic growth in maturing dentate granule cells ([@B27]), suggesting that NMDA hypofunction in the DG may affect dendritic arborization in this brain region that exhibits developmental regulated changes in neurogenic activity ([@B21]).
In the current study we sought to determine if the reduction in NMDAR function observed in *Fmr1* KO animals was associated with developmental deficits in dendritic arborization of neurons in the hippocampal dentate granule cell layer (GCL) ([@B8]). The adult-born neurons of the DG have been shown to express primarily GluN2B subunits early on ([@B42]), but are also known to undergo extensive dendritic arborization as they migrate into an already extensively populated GCL ([@B46]). Indeed, dendritic arborization and cell body positioning have been used to identify young and old neurons in the DG ([@B49]; [@B16]). Newly generated neurons tend to be preferentially located in the inner layer of the GCL ([@B35]; [@B17]; [@B36]), whereas more mature granule cells appear to be located in the outer GCL ([@B49]; [@B35]; [@B36]). Combined morphological and electrophysiological analyses also indicate that neurons in the outer GCL are morphologically more complex and thus have a lower series resistance than neurons in the inner GCL ([@B49]; [@B46]; [@B27]). Using the location of neurons in the GCL as a means to select neurons for whole-cell patch clamp analyses, we investigated how the loss of FMRP affects NMDAR function and dendritic arborization of both younger and more mature hippocampal DG neurons.
Materials and Methods {#s1}
=====================
Animals
-------
Adult male *Fmr1* KO mice with a C57BL/6 genetic background ([@B2]) and their wild-type (WT) littermates at the age of 4- to 6-week month old were used for the experiments. All mice housed with food and water available *ad libitum* on a 12 h light/dark cycle. All experiments were performed in accordance with the guidelines set out by the Canadian Council on Animal Care and approved by the University of Victoria Animal Care Committee.
Electrophysiology
-----------------
### Electrophysiological Preparation
Adult mice were anesthetized with isoflurane, their brains removed, and transverse hippocampal slices were prepared as previously described ([@B47]). Briefly, hippocampal slices (350 μm) were acquired using a Vibratome 1500 (Ted Pella, Inc., Redding, CA, United States). The brain was immersed in oxygenated (95% O~2~/5% CO~2~) artificial cerebrospinal fluid (ACSF) containing (in mM) 125 NaCl, 3 KCl, 1.25 NaHPO~4~, 25 NaHCO~3~, 1 CaCl~2~, 6 MgCl~2~, and 25 Glucose at 4°C. After sectioning, slices were transferred to a holding chamber containing warm (30°C) oxygenated normal ACSF (nACSF) consisting of (in mM) 125 NaCl, 2.5 KCl, 1.25 NaHPO~4~, 25 NaHCO~3~, 2 CaCl~2~, 1.3 MgCl~2~, and 10 dextrose for 1 h before being used for electrophysiological recordings.
### Whole Cell Recording
Cells were patched using a borosilicate glass recording electrode (5--7 MΩ) and the formation of a gigaseal (2 GΩ) was required prior to break-in. Recordings with a series resistance higher than 30 MΩ or presenting a variation of more than 10% were excluded from the analyses. The intracellular solution consisted of (in mM) 20 KCl, 120 [K]{.smallcaps}-gluconate, 4 NaCl, 0.1 EGTA, 4 ATP, 0.3 GTP, 14 Phosphocreatine (Osmolarity 270 mOsm/kg, pH 7.2) when action potentials were measured in current clamp mode. To examine NMDA/AMPA receptor mediated excitatory post-synaptic currents (EPSCs) in Voltage-Clamp mode, the internal solution was composed of (in mM) 135 Cesium methanesulfonate, 8 NaCl, 10 HEPES, 2 ATP, 0.3 GTP, 7 Phosphocreatine, 10 QX-314 (Osmolarity 280 mOsm/kg, pH 7.3) and biocytin (0.2--0.4%). In all cells, Alexa Fluor 488 (40 mM) was included in the intracellular solution to assist with the visualization and classification of granule cells. EPSCs were evoked with bipolar stimulating electrodes placed in the medial perforant pathway and recorded using Axopatch 200B amplifier and pClamp10 software (Axon Instruments). AMPAR-mediated EPSCs were measured at a holding voltage of -70 mV, while NMDA EPSCs were measured by applying a +40 mV holding potential in the presence of picrotoxin (100 μM) in nACSF. Some granule cells located in the inner cell layer did not display NMDA receptor currents; however, only granules cells showing both NMDAR and AMPAR EPSCs were included in the analyses.
Intracellular Filling and Immunostaining of Biocytin-Filled Cells
-----------------------------------------------------------------
Cells were filled with Alexa Fluor 488 together with biocytin for immediate visualization with fluorescence microscopy (Olympus Fluoview 1000). After each recording, the electrode was quickly retracted from the cell to help maintain cell integrity for histology. Slices were then fixed with 4% paraformaldehyde and left overnight at 4°C. The following day they were washed with 0.01 M PBS repeatedly and then incubated in 3% H~2~O~2~ for 45 min to block any endogenous peroxidase activity. Slices were then washed with 0.01 M PBS, before being incubated in an avidin-biotinylated HRP complex (ABC) solution containing 0.1% Triton-X for 48 h at room temperature. Biocytin-filled cells were visualized using a diaminobenzidine (DAB; Sigma-Aldrich) solution. Slices were then mounted onto gelatin-coated glass slides and allowed to dry at room temperature for 2 days before being dehydrated in graded ethanol and cover-slipped with Permount.
Co-labeling of Biocytin-Filled Cells With Neuronal Markers
----------------------------------------------------------
Following antigen retrieval in citric acid buffer for 15 min (pH 6.8 at 60°C), brain slices were washed thoroughly with 0.01 M PBS and then incubated with primary antibodies: rabbit anti-doublecortin (Abcam, 1: 200, Cat No.: ab18723) or mouse anti-NeuN (Millipore, 1: 200, Cat No.: MAB377) at 4°C for 3 days. Brain slices were then washed with PBS and incubated in Streptavidin-conjugated with Cy3 (Sigma, 1: 400 Cat No.: 6402) and Alexa Fluor 488 conjugated donkey anti-rabbit or donkey anti-mouse IgG antibodies (Life Technologies, 1: 200, Cat No.: S-11223) for 4 h at room temperature. Following washes in PBS, slides were coverslipped using Fluoromount (Thermo Fisher Scientific).
Selection and Classification of Granule Cells
---------------------------------------------
We and others have previously shown that granule cells located in the outer GCL tend to have several primary dendrites and more dendritic branching, while granule cells located in the inner GCL are more likely to have only one primary dendrite and less dendritic branching ([@B13]; [@B22]; [@B9]; [@B49]; [@B27]). In accordance with our prior work, young and mature dentate granule neurons were selected based on their position in the inner and outer GCL, as well as their morphology, respectively. These cells typically had either single (inner cells) or multiple (outer cells) primary dendrites extending from the cell body, as previously reported ([@B49]; [@B3]; [@B27]; [@B53]).
Sholl Analysis of Dendritic Complexity
--------------------------------------
Only those granule cells that exhibited intact dendrites with no cut branches were used in these analyses. Dendritic tracing was performed using Neurolucida software (MBF Bioscience, Williston, VT, United States) with a 40X objective lens. Sholl analysis was used to measure dendritic lengthen and dendritic branching with a concentric 10-μm interval as previously reported ([@B27]; [@B53]).
Statistical Analysis
--------------------
Repeated measure ANOVA for dendritic branching and dendritic length analysis were performed using SPSS 14.0 (SPSS, Inc., Chicago, IL, United States). In some instances, two group comparisons between WT and *Fmr1* KO were performed using Student's *t*-test. Data are presented as mean ± SEM. Statistical significance was indicated by a probability (*P*) value less than 0.05.
Results
=======
Morphological Difference of Biocytin-Filled Granule Cells Located in the Inner and Outer Granule Cell Layers
------------------------------------------------------------------------------------------------------------
The neurons co-labeled with the immature neuronal marker doublecortin (DCX) displayed as the ones with a single primary dendrite were located in the inner cell layer (Figures [1A](#F1){ref-type="fig"}--[C](#F1){ref-type="fig"}), while neurons co-labeled with the mature neuronal marker NeuN displayed multiple dendrites and were located in the outer cell layer (Figures [1D](#F1){ref-type="fig"}--[F](#F1){ref-type="fig"}). Representative images showing differences in morphology and location of granule neurons in the GCL are demonstrated in Figures [1G](#F1){ref-type="fig"}--[I](#F1){ref-type="fig"}.
{#F1}
Action Potential and Membrane Properties of Dentate Granule Cells in *Fmr1* KO Mice
-----------------------------------------------------------------------------------
Figure [2A](#F2){ref-type="fig"} shows a representation of action potential trains in neurons with a single primary dendrite or multiple primary dendrites from WT mice. There were no significant differences in action potential frequency when comparing granule cells from WT to *Fmr1* KO animals (single primary dendrite: *F*~1,20~ = 0.007, *P* = 0.933, Figure [2B](#F2){ref-type="fig"}; multiple primary dendrites: *F*~1,24~ = 1.158, *P* = 0.293, Figure [2C](#F2){ref-type="fig"}). However, granule neurons with multiple primary dendrites from *Fmr1* KO mice displayed a trend toward a higher maximum frequency of action potential (Student's *t*-test, WT: *P* = 0.156; *Fmr1* KO: *P* = 0.05, Figure [2D](#F2){ref-type="fig"}), and a trend toward a lower input resistance when compared with the ones with a single primary dendrite (Table [1](#T1){ref-type="table"}).
{#F2}
######
Membrane properties of granule cells with single and multiple primary dendrites that display both AMPAR and NMDAR-EPSCs.
WT *Fmr1* KO
---------- --------------- --------------- --------------- --------------- --------
Ri (mΩ) 344.99 ± 15.1 223.45 ± 8.92 299.21 ± 32.4 211.09 ± 16.7 \>0.05
RMP (mV) -74.06 ± 1.03 -75.9 ± 0.60 -72.44 ± 3.73 -72.68 ± 1.12 \>0.05
Decreased NMDA EPSCs in Hippocampal Granule Cells From *Fmr1* KO Mice
---------------------------------------------------------------------
For all recordings, AMPAR and NMDAR EPSCs were evoked using increasing stimulation intensity to construct I/O curves and determine the maximum response size. Representative traces acquired from granule neurons are depicted in Figure [3A](#F3){ref-type="fig"}. Similar levels of AMPAR-mediated EPSCs were recorded in younger granule cells from both experimental groups (*F*~1,25~ = 1.36, *P* = 0.254, Figure [3B](#F3){ref-type="fig"}), while a trend toward a reduction in NMDAR response size was observed in the *Fmr1* KO mice when compared to their WT littermates (*F*~1,25~ = 3.69, *P* = 0.067, Figure [3C](#F3){ref-type="fig"}). No alterations were found in AMPAR-mediated EPSCs of mature cells (*F*~1,20~ = 1.82, *P* = 0.194, Figure [3D](#F3){ref-type="fig"}), but a significant reduction in NMDAR-mediated EPSCs was observed in mature granule cells from *Fmr1* KO mice (*F*~1,20~ = 14.843, *P* = 0.001, Figure [3E](#F3){ref-type="fig"}).
{#F3}
The analyses of the maximum amplitude of AMPAR-mediated EPSCs in neurons with a single primary dendrite did not reveal a significant difference between genotypes (*P* = 0.89), but a trend toward a reduction in NMDAR EPSCs was observed in cells from *Fmr1* KO mice in comparison with WT mice (*P* = 0.09, Figure [4A](#F4){ref-type="fig"}). This trend was also observed in the AMPAR/NMDAR ratio (*P* = 0.07, Figure [4B](#F4){ref-type="fig"}). A significant decrease in absolute maximum NMDAR-mediated EPSCs was found in mature neurons from *Fmr1* KO mice (*P* \< 0.01, Figure [4C](#F4){ref-type="fig"}), as well as an increase in the AMPAR/NMDAR ratio (*P* \< 0.05, Figure [4D](#F4){ref-type="fig"}) when compared to their WT counterparts. The amplitude of individual AMPAR- and NMDAR-mediated responses from hippocampal neurons at distinct developmental stages is depicted in Figure [4E](#F4){ref-type="fig"}.
{#F4}
*Fmr1* KO Mice Present Decreased Dendritic Complexity in More Mature Granule Cells, but Not Younger Granule Cells
-----------------------------------------------------------------------------------------------------------------
Examples of biocytin filled cells used for the analysis of dendritic complexity are shown in Figures [5A,B](#F5){ref-type="fig"}. Sholl analysis revealed that WT and *Fmr1* KO mice presented similar dendritic measures (dendritic length: *F*~1,18~ **=** 0.185, *P* = 0.673, Figure [5C](#F5){ref-type="fig"}; dendritic branching: *F*~1,18~ = 0.473, *P* = 0.501, Figure [5D](#F5){ref-type="fig"}). Conversely, cells with multiple primary dendrites from *Fmr1* KO mice displayed a drastic reduction in both dendritic length (*F*~1,22~ **=** 26.291, *P* \< 0.005, Figure [5E](#F5){ref-type="fig"}) and number of intersections (*F*~1,22~ = 15.301, *P* = 0.001, Figure [5F](#F5){ref-type="fig"}) when compared to their WT littermates. These findings indicate that the loss of FMRP may lead to a significant decrease in the development of dendritic complexity in more mature neurons.
{#F5}
Discussion
==========
The current study indicates that a loss of FMRP results in a significant reduction in NMDAR-mediated EPSCs and a concomitant decrease in dendritic complexity in hippocampal granule cells. Our findings also indicate that these deficits are primarily observed in the mature granule neurons which are with multiple primary dendritic processes arising from the soma, and are primarily located in the outer GCL. On the other hand, younger granule neurons that reside in the inner GCL, did not show a significant change in dendritic complexity, although they did show a trend toward a reduction in NMDAR EPSCs induced by the lack of FMRP. These data further support our previous findings indicating that *Fmr1* KO mice present a dysfunction in NMDAR in the hippocampal DG ([@B6]; [@B52]), and show for the first time that these deficits are preferentially associated with a reduction in dendritic arborization in a specific population of granule cells in the outer GCL.
Dendritic morphology can affect synaptic plasticity by modulating neuronal firing rates, as well as altering the propagation of EPSPs and action potentials ([@B31]; [@B30]; [@B48]). Loss of FMRP has been shown to reduce pruning of dendrites in the somatosensory cortex ([@B19]) and in mitral cells of the olfactory bulb ([@B20]), suggesting that this protein plays an important role in the development of dendritic processes. In the current study, we found that the loss of FMRP leads to a significant decrease in dendritic complexity in a specific subpopulation of granule neurons located in the outer GCL. This population of newborn neurons present multiple primary dendrites arising from the cell body and has been previously classified as being "mature neurons" based on morphological and electrophysiological analyses ([@B28]; [@B49]; [@B3]; [@B27]). Our findings are in line with previous studies showing that the lack of FMRP leads to significant impairments in neurite extension of primary neural progenitor cells derived from the hippocampal DG ([@B23]). In addition, an *in vivo* study using retroviral labeling with green fluorescent protein found that mature newborn neurons from *Fmr1* KO mice present a decrease in dendritic complexity when compared to their WT counterparts ([@B24]). Our study corroborate these findings and provide further evidence indicating that the lack of FMRP affects hippocampal newborn cells in an age-dependent manner.
The regulation of neuronal morphology and synaptic modifications depends on the remodeling of the actin cytoskeleton, which in turn has been previously shown to be regulated by FMRP ([@B14]; [@B34]). Thus, it is possible that the deficits in dendritic length and branching observed in the present study are related to alterations in the organization of actin filaments induced by the lack of FMRP in the hippocampal DG. These changes may be related to the NMDAR hypofunction we observed, since these receptors are known to play a critical role in activity-dependent development of dendritic arbors and synaptogenesis ([@B33]; [@B51]). It may be that the deficits in NMDAR function observed in our study in turn affect several NMDAR-dependent proteins who exert an influence on dendritic morphology. These may include extracellular signal-regulated kinases ([@B45]), Ca^2+^/calmodulin-dependent protein kinase II ([@B41]), Rho family GTPases ([@B44]) and glycogen synthase kinase 3 β (GSK-3β) ([@B38]). These proteins are interesting targets for future studies, particularly GSK-3β, since alterations in the translation of this protein seem to play a major role in the impaired differentiation of hippocampal neurons in FXS ([@B29]; [@B24]).
Prior research indicates that GluN2A subunit-containing NMDARs are important for dendritic arborization ([@B27]), while GluN2B may be more critical in regulating spine formation ([@B25]). Interestingly, we have previously shown that NMDAR hypofunction is associated with a significant decrease in the expression of both GluN2A and GluN2B subunits in the DG of *Fmr1* KO mice ([@B6]). However, because FMRP is a translational repressor, it is still not clear how its deletion leads to a reduction in the expression of NMDAR subunits, as well as to what extent it affects temporal changes in GluN2A and GluN2B expression in younger and more mature dentate granule cells. One factor that could be contributing to this deficit is the dysregulation of translational responses induced by excessive mGluR activation ([@B32]), disruption of PSD-95 and CAMKII activities ([@B55]) and/or by the abnormal morphology of dendritic spines seen in FXS ([@B10]; [@B4]). On the other hand, since the loss of FMRP has an impact in the activity of a vast number of proteins, we cannot rule out the possibility that NMDAR-independent mechanisms are also contributing to the alterations in dendritic development found in our study. Regardless, the fact that we did not observe significant reductions in dendritic complexity and NMDAR function in the younger population of granule cells, suggests that the effects of FMRP on NMDAR function and dendritic complexity may occur over time. The DG offers a unique opportunity to study developmental changes, as it is one of the brain areas that exhibits continual neurogenesis throughout the lifespan. Indeed, the loss of FMRP also alters hippocampal neurogenesis in adult animals, which show increased cell proliferation, but impaired neuronal differentiation ([@B29]; [@B24]). Together these findings may suggest that the system is trying to compensate for reduced synaptic signaling by enhancing the production of new cells, but that loss of FMRP also negatively impacts the development of these cells. Our data indicate that a loss of FMRP causes significant deficits in both NMDAR function and dendritic arborization in mature neurons, and this could contribute to the abberant neurogenic process and reduced cognitive performance that has been observed with the loss of FMRP.
Author Contributions
====================
All authors made substantial contributions to the conception or design of the work; or the acquisition, analysis or interpretation of data for the work. Drafting the work or revising it critically for important intellectual content. Provide approval for publication of the content. Agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
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.
**Funding.** This project was supported by grants from the Canadian Institutes for Health Research (CIHR MOP 125888) and NSERC awarded to BC. S-YY was supported by a postdoctoral fellowship from the CIHR and the Fragile X Research Foundation of Canada. LB was supported by IBRO and Michael Smith Foundation for Health Research (MSFHR) postdoctoral fellowships.
[^1]: Edited by: Regina Dahlhaus, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany
[^2]: Reviewed by: Christina Gross, Cincinnati Children's Hospital Medical Center, United States; Michael E. Cahill, University of Wisconsin-Madison, United States; Se-Young Choi, Seoul National University, South Korea
[^3]: ^†^Co-first authors
[^4]: ^‡^Present address: Suk-Yu Yau, Department of Rehabilitation Sciences, Hong Kong Polytechnic University, Hong Kong, Hong Kong
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .111}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#section5-2050312120915399}
============
Collaboration within medical research accelerates innovation and promotes complex scientific endeavors. Technological advances facilitating communication and the ability to record and share data have simplified the collaborative process, revolutionizing contemporary research practices. Not surprisingly, research has grown in both scale and rigor. Today, multi-disciplinary and multi-institutional collaborations are the norm and are often considered necessary for maintaining and achieving the highest clinical research standards.
This shift has been accompanied by a growth in the number of authors per scientific manuscript. Single and few-authored papers, common in decades prior, are exceedingly rare today. This is especially true in publications with original data. Curiously, these trends appear to be pervasive in nearly all-academic fields. *The Economist* recently observed increasing authorship trends in various academic disciplines, including Arts & Humanities, Economics, Engineering, Chemistry, Medicine, and Physics & Astronomy. Notably, medicine had the second greatest rate of authorship growth, behind publications in Physics & Astronomy.^[@bibr1-2050312120915399]^ Increasing authorship has been independently observed in Pediatric Surgery,^[@bibr2-2050312120915399]^ Radiology,^[@bibr3-2050312120915399]^ Radiation Oncology,^[@bibr4-2050312120915399]^ Neurology,^[@bibr5-2050312120915399]^ Urology,^[@bibr6-2050312120915399]^ Sports Medicine,^[@bibr7-2050312120915399]^ Orthopedics,^[@bibr8-2050312120915399]^ Psychiatry,^[@bibr9-2050312120915399]^ and Obstetrics and Gynecology (OB/GYN)^[@bibr10-2050312120915399]^ literature, but without systematic comparisons.
It has been proposed that the increased authorship stems from the need for greater manpower to conduct high-quality research endeavors.^[@bibr11-2050312120915399]^ Others attribute increased authorship to widespread authorship inflation driven by growing academic pressures, decreased funding resources, and variable measures for academic promotion across fields and institutions.^[@bibr10-2050312120915399],[@bibr12-2050312120915399],[@bibr13-2050312120915399]^ The exact driving forces behind this trend are uncertain and likely multifactorial.
Traditionally, academic credit is awarded to individuals who make "sufficient" contributions in the form of authorship. The International Committee of Medical Journal Editors (ICMJE)^[@bibr14-2050312120915399]^ defined authorship as fulfilling the following four criteria: (1) substantial contributions to the conception or analysis of data, (2) drafting the work or revising it critically for important intellectual content, (3) final approval of submitted manuscript, and (4) agreement to be accountable for all aspects of the work. Contributors who do not fulfill these four criteria should be credited through an acknowledgment. However, in situations where many collaborators have contributed in varying degrees and forms, it can be difficult to discern whose contributions justify authorship and to what capacity. In this study, we aim to describe authorship trends within contemporary medical literature and to assess how these patterns compare across different medical specialties. Ultimately, the goal of this research is to elucidate the driving forces behind this change and promote the implementation of standards that guide authors to appropriately assign authorship.
Methods {#section6-2050312120915399}
=======
Specialty, journals, and article information {#section7-2050312120915399}
--------------------------------------------
The Association of American Medical Colleges^[@bibr15-2050312120915399]^ (AAMC, Washington, DC) Careers in Medicine^®^ tool and the Report on Residents^®^ were used to identify 23 medical specialties in which training is entered directly by graduating senior medical students.^[@bibr16-2050312120915399]^ Information including years of residency training, percentage of recent graduates with academic faculty positions, and surgical versus non-surgical specialty were collected. The 2016 Journal Citation Reports^®[@bibr17-2050312120915399]^ (Thomson Reuter, New York) was used to identify three of the highest impact-factored, clinically orientated, specialty-specific publications for each of the 23 specialties. The National Institutes of Health's (NIH) National Library of Medicine PubMed database was used to obtain PubMed IDs (PMID) of case reports, review articles, and original research published between 1 January 2005 and 31 December 2017 in the 69 selected journals included in this study. We excluded all articles listed under addresses, biographies, comment, dataset, editorials, abstracts, guidelines, personal narratives, erratum, and video--audio media.
The NIH Office of Portfolio Analysis's web-based next-generation portfolio analysis platform, iCite, was queried for bibliometric data including article title, author list, year of publication, article type, journal in which it was published, and relative citation ratio (RCR, a field-normalized metric which shows the scientific influence of each article relative to the average NIH-funded article in the same discipline published in the same year) for each article selected for analysis.^[@bibr18-2050312120915399]^ Permission and assistance with use of the iCite program were obtained from the program developers in Bethesda, Maryland, in 2017 when this study was first conceptualized. Article type was categorized strictly based on how they were recorded in the PubMed database to maintain consistency between classification between journals. This observational study did not involve human subjects and was therefore exempt from institutional ethics review.
Statistical analysis {#section8-2050312120915399}
--------------------
Statistical analyses were performed on Stata/SE 14 (StataCorp LLC, College Station, Texas), with two-sided p-values \< 0.05 predetermined as the threshold for statistical significance. Article citation characteristics were described by article type, by specialty, and for all articles. One-way analysis of variance (ANOVA) compared authors per article, number of total citations, citations per year, and RCR by article type. Multivariable linear regression analysis of non-collinear variables (article year, specialty, article type) as predictors of authors per article was performed and used to calculate predicted values. Stratified regression models were repeated independently within each article type (with specialty as co-variable) and within each specialty (with article type as co-variable). Adjusted coefficients for publication year on authors per article were reported; analyses were performed for all articles, by article type, and by specialty. Predicted values from the three article-type stratified regression models were used to produce a tabular and several graphical illustrations. Specialty characteristics including surgical versus non-surgical specialty type, minimum years of residency training, and percent of graduates with current academic positions were evaluated using multivariable regression as potential predictors of mean authors per specialty article. Ninety-five percent confidence interval (CI) and standard errors of coefficients were calculated using the bootstrap sampling with 50 replications.
Results {#section9-2050312120915399}
=======
Summary of specialty, journal, and articles evaluated {#section10-2050312120915399}
-----------------------------------------------------
Overall, 121,397 peer-reviewed publications were evaluated---of which, 45.1% were original research, 28.7% review articles, and 26.3% case reports ([Table 1](#table1-2050312120915399){ref-type="table"}). Original research had significantly more authors per article (7.26 authors, adjusted mean), than review articles (4.63 authors, adjusted mean) and case reports (4.46 authors, adjusted mean) (ANOVA: p \< 0.001). Average RCR, metric for article impact, was similarly greatest in review articles (RCR: 2.98), followed by original research (RCR: 2.44) and case report (RCR: 0.54) (ANOVA: p \< 0.001) ([Table 1](#table1-2050312120915399){ref-type="table"}). Each consecutive year was associated with an increase of 0.16 authors per publication for all articles and specialties combined (p \< 0.001) ([Table 2](#table2-2050312120915399){ref-type="table"}). Original research articles were associated with the greatest authorship growth (growth of 0.23 authors/year, an increase from 5.87 to 8.51 authors between 2005 and 2017), as compared with review articles (0.18 authors/year, 3.53 to 5.69 authors) and case reports (0.01 authors/year, 4.26 to 4.49 authors) (p \< 0.001 for all) ([Figure 1](#fig1-2050312120915399){ref-type="fig"}, [Table 2](#table2-2050312120915399){ref-type="table"}). Case reports initially had more authors than review articles, but this reversed over time---with review articles later having more authors than case reports by about 2010 ([Table 3](#table3-2050312120915399){ref-type="table"}, [Figure 1](#fig1-2050312120915399){ref-type="fig"}). This trend was also seen in the majority of evaluated specialties ([Figure 3](#fig3-2050312120915399){ref-type="fig"}).
######
Summary of articles, citations, and RCR by specialty evaluated.
Citations Citations per year RCR
------------------------------------------------------ -------------- -------------------- ----------------
By article type
Case report (n = 31,877) 3 (1--8) 0.4 (0.1--1.0) 0.3 (0.1--0.6)
Review article (n = 34,799) 15 (4--40) 2.7 (1.0--6.2) 1.6 (0.6--3.5)
Original research (n = 54,721) 18 (7--38) 2.7 (1.3--5.2) 1.5 (0.8--2.9)
By specialty
Anesthesiology (n = 5515) 14 (5--34) 2 (0.7--4.7) 1.2 (0.4--2.7)
*Anaesthesia*: n = 1658 (30.1%)
*Anesthesiology*: n = 1856 (33.7%)
*Br J Anaesth*: n = 2001 (36.3%)
Dermatology (n = 5683) 7 (2--21) 1.3 (0.4--3.6) 0.8 (0.3--1.9)
*J Am Acad Dermatol*: n = 3858 (67.9%)
*J Invest Dermatol*: n = 1014 (17.8%)
*JAMA Dermatol*: n = 811 (14.3%)
Emergency Medicine (n = 3129) 7 (1--21) 1.3 (0.2--3.0) 0.8 (0.2--1.8)
*Acad Emerg Med*: n = 1100 (35.2%)
*Ann Emerg Med*: n = 1656 (52.9%)
*Scand J Trauma Resusc Emerg Med*: n = 373 (11.9%)
Family Medicine (n = 1234) 11 (3--28) 1.7 (0.7--3.4) 1.0 (0.4--1.9)
*Ann Fam Med*: n = 318 (25.8%)
*Br J Gen Pract*: n = 462 (37.4%)
*J Am Board Fam Med*: n = 454 (36.8%)
General Surgery (n = 3514) 22 (7--50) 4.6 (2.0--8.5) 2.6 (1.2--4.7)
*Ann Surg*: n = 1379 (39.2%)
*Br J Surg*: n = 1683 (47.9%)
*JAMA Surg*: n = 452 (12.9%)
Internal Medicine (n = 3524) 28 (7--78) 5.5 (1.7--12.5) 2.5 (0.8--5.5)
*Ann Intern Med*: n = 2029 (57.6%)
*J Intern Med*: n = 836 (23.7%)
*JAMA Intern Med*: n = 659 (18.7%)
Interventional Radiology (n = 4247) 4 (1--11) 0.7 (0.2--1.8) 0.5 (0.1--1.1)
*Cardiovasc Intervent Radiol*: n = 1627 (38.3%)
*J Vasc Interv Radiol*: n = 2052 (48.3%)
*Semin Intervent Radiol*: n = 568 (13.4%)
Neurology (n = 2594) 48 (15--109) 8 (3.5--15.5) 3.5 (1.6--6.6)
*Brain*: n = 899 (34.7%)
*JAMA Neurol*: n = 529 (20.4%)
*Lancet Neurol*: n = 1166 (44.9%)
Neurosurgery (n = 7569) 8 (2--21) 1.3 (0.5--3.0) 1.0 (0.4--2.0)
*J Neurosurg*: n = 2121 (28.0%)
*Neurosurgery*: n = 2976 (39.3%)
*World Neurosurg*: n = 2472 (32.7%)
Obstetrics & Gynecology (n = 6743) 15 (5--35) 2.1 (0.8--4.7) 1.2 (0.5--2.6)
*Am J Obstet Gynecol*: n = 2489 (36.9%)
*Hum Reprod*: n = 1366 (20.3%)
*Obstet Gynecol*: n = 2888 (42.8%)
Ophthalmology (n = 5267) 14 (4--34) 2.3 (0.8--5.0) 1.5 (0.6--3.2)
*Am J Ophthalmol*: n = 1869 (35.5%)
*JAMA Ophthalmol*: n = 814 (15.5%)
*Ophthalmology*: n = 2584 (49.1%)
Orthopedic Surgery (n = 5841) 15 (5--34) 2.2 (1.0--4.7) 1.6 (0.7--3.2)
*Clin Orthop Relat Res*: n = 2496 (42.7%)
*J Am Acad Orthop Surg*: n = 996 (17.1%)
*J Bone Joint Surg Am*: n = 2349 (40.2%)
Otolaryngology (n = 3562) 5 (1--13) 0.8 (0.2--1.9) 0.6 (0.2--1.4)
*Clin Otolaryngol*: n = 567 (15.9%)
*JAMA Otolaryngol Head Neck Surg*: n = 500 (14.0%)
*Otolaryngol Head Neck Surg*: n = 2495 (70.0%)
Pathology (n = 2073) 24 (10--49) 3.2 (10--49) 1.5 (0.8--2.7)
*Am J Surg Pathol*: n = 959 (46.3%)
*J Pathol*: n = 505 (24.4%)
*Mod Pathol*: n = 609 (29.4%)
Pediatrics (n = 7977) 16 (5--38) 2.7 (1.0--5.5) 1.4 (0.6--3.0)
*J Pediatr*: n = 2522 (31.6%)
*JAMA Pediatr*: n = 404 (5.1%)
*Pediatrics*: n = 5051 (63.3%)
Physical Medicine & Rehabilitation (n = 3981) 9 (3--22) 1.6 (0.6--3.3) 1.1 (0.5--2.2)
*Am J Phys Med Rehabil*: n = 1256 (31.5%)
*Arch Phys Med Rehabil*: n = 2234 (56.1%)
*Eur J Phys Rehabil Med*: n = 491 (12.3%)
Plastic Surgery (n = 6174) 6 (2--15) 0.8 (0.2--15) 0.7 (0.2--1.8)
*J Plast Reconstr Aesthet Surg*: n = 2521 (40.8%)
*JAMA Facial Plast Surg*: n = 137 (2.2%)
*Plast Reconstr Surg*: n = 3516 (56.9%)
Psychiatry (n = 2929) 38 (14--85) 5.3 (2.2--10.5) 2.8 (1.3--5.2)
*Am J Psychiatry*: n = 1776 (60.6%)
*Br J Psychiatry*: n = 849 (29.0%)
*JAMA Psychiatry*: n = 304 (10.4%)
Radiation Oncology (n = 4580) 19 (8--40) 3.0 (1.5--5.4) 1.4 (0.7--2.5)
*Int J Radiat Oncol Biol Phys*: n = 2709 (59.1%)
*Radiother Oncol*: n = 1443 (31.5%)
*Semin Radiat Oncol*: n = 428 (9.3%)
Radiology (n = 5809) 13 (4--32) 1.8 (0.7--4.1) 1.1 (0.5--2.4)
*AJR Am J Roentgenol*: n = 3166 (54.5%)
*J Am Coll Radiol*: n = 701 (12.1%)
*Radiology*: n = 1942 (33.4%)
Thoracic Surgery (n = 12,848) 5 (1--16) 0.8 (0.2--2.5) 0.5 (0.1--1.3)
*Ann Thorac Surg*: n = 7037 (54.8%)
*Eur J Cardiothorac Surg*: n = 2684 (20.9%)
*J Thorac Cardiovasc Surg*: n = 3127 (24.3%)
Urology (n = 8932) 15 (4--36) 2.1 (0.7--5.0) 1.1 (0.4--2.5)
*Eur Urol*: n = 2056 (23.0%)
*J Urol*: n = 3317 (37.1%)
*Urology*: n = 3559 (39.8%)
Vascular Surgery (n = 7672) 7 (2--19) 1.1 (0.4--3.0) 0.7 (0.3--1.6)
*Ann Vasc Surg*: n = 2345 (30.6%)
*Eur J Vasc Endovasc Surg*: n = 1654 (21.6%)
*Vasc Surg*: n = 1654 (21.6%)
Overall
Total (n = 121,397) 11 (3--30) 1.8 (0.5--4.3) 1.1 (0.4--2.5)
RCR: relative citation ratio; ANOVA: analysis of variance.
Citation count, citations per year, and RCR were compared by article type and by specialty using the one-way ANOVA test. Journals names written in PubMed MedAbbr format. All values presented as median (interquartile range) unless otherwise specified.
######
Multivariable linear regression analysis of non-collinear variables (specialty, article type, and year) as predictors of authors per article.
Coefficient p-value 95% CI
------------------------------------------------ ------------- --------- ------------------
By article type
Case report (n = 31,877) (Reference)
Review article (n = 34,799) 0.04 0.119 \[--0.01, 0.1\]
Original research (n = 54,721) 2.67 \<0.001 \[2.62, 2.72\]
By specialty
Orthopedic Surgery (n = 5841) (Reference)
Otolaryngology (n = 3562) 0.02 0.669 \[--0.09, 0.13\]
Physical Medicine & Rehabilitation (n = 3981) 0.11 0.074 \[--0.01, 0.22\]
Plastic Surgery (n = 6174) 0.12 0.003 \[0.04, 0.19\]
Family Medicine (n = 1234) 0.13 0.110 \[--0.03, 0.29\]
Anesthesiology (n = 5515) 0.45 \<0.001 \[0.35, 0.55\]
Emergency Medicine (n = 3129) 0.49 \<0.001 \[0.35, 0.62\]
Obstetrics & Gynecology (n = 6743) 0.84 \<0.001 \[0.75, 0.93\]
Vascular Surgery (n = 7672) 0.90 \<0.001 \[0.81, 0.99\]
Ophthalmology (n = 5267) 1.10 \<0.001 \[0.98, 1.22\]
Interventional Radiology (n = 4247) 1.25 \<0.001 \[1.12, 1.39\]
Pediatrics (n = 7977) 1.25 \<0.001 \[1.15, 1.36\]
Dermatology (n = 5683) 1.28 \<0.001 \[1.16, 1.41\]
Radiology (n = 5809) 1.30 \<0.001 \[1.19, 1.41\]
Neurosurgery (n = 7569) 1.41 \<0.001 \[1.32, 1.49\]
Thoracic Surgery (n = 12,848) 1.64 \<0.001 \[1.56, 1.71\]
General Surgery (n = 3514) 1.71 \<0.001 \[1.56, 1.85\]
Urology (n = 8932) 1.71 \<0.001 \[1.6, 1.82\]
Internal Medicine (n = 3524) 1.83 \<0.001 \[1.64, 2.02\]
Psychiatry (n = 2929) 2.19 \<0.001 \[1.96, 2.42\]
Pathology (n = 2073) 2.36 \<0.001 \[2.16, 2.57\]
Radiation Oncology (n = 4580) 2.97 \<0.001 \[2.83, 3.11\]
Neurology (n = 2594) 4.16 \<0.001 \[3.80, 4.52\]
By Year
Year 0.16 \<0.001 \[0.15, 0.16\]
CI: confidence interval.
Variables are ordered from smallest to largest coefficient values.
{#fig1-2050312120915399}
######
Authors per article by year, predicted from article-type stratified regression models.
Case report Review article Original research
------------------------------ ------------- ---------------- ------------------- ------ ------ ------
Authors per article, by year
2005 4.26 0.01 3.53 0.01 5.87 0.01
2006 4.27 0.01 3.67 0.01 6.03 0.01
2007 4.29 0.01 3.87 0.01 6.31 0.01
2008 4.32 0.01 4.01 0.01 6.48 0.01
2009 4.33 0.01 4.22 0.01 6.71 0.01
2010 4.38 0.01 4.39 0.01 6.93 0.02
2011 4.35 0.01 4.59 0.01 7.09 0.01
2012 4.37 0.01 4.78 0.01 7.32 0.01
2013 4.43 0.01 4.98 0.01 7.68 0.01
2014 4.45 0.01 5.16 0.01 7.91 0.01
2015 4.42 0.01 5.32 0.01 8.11 0.01
2016 4.50 0.01 5.53 0.01 8.36 0.01
2017 4.49 0.01 5.69 0.01 8.51 0.01
Authors per article, overall
2005--2017 4.46 0.00 4.63 0.00 7.26 0.00
SE: standard error.
Predicted values were obtained from the linear regression model fitted and adjusted for authorship year, article type, and specialty.
Authorship trends by specialty {#section11-2050312120915399}
------------------------------
Specialties with the greatest number of authors per article were as follows: Neurology, Radiation Oncology, Pathology, Psychiatry, and Internal Medicine (from greatest to less). Specialties with the fewest authors per article were as follows: Orthopedic Surgery, Otolaryngology, Physical Medicine and Rehabilitation, and Plastic Surgery. Orthopedic Surgery had the fewest adjusted authors per article, whereas Neurology had the most---4.16 more authors than Orthopedic Surgery ([Table 2](#table2-2050312120915399){ref-type="table"}). The majority (22/23) of all specialties independently experienced statistically significant adjusted authorship growth for all articles (p \< 0.001 for these 22 specialties). Interventional Radiology was the only specialty that did not show a statistically significant authorship growth or decline (p = 0.12). Specialties with the greatest authorship growth were Neurology (growth of 0.42 authors/year), Psychiatry (0.35 authors/year), General Surgery (0.29 authors/year), Urology (0.27 authors/year), and Pathology (0.27 authors/year). Specialties with the least authorship growth were Vascular Surgery (0.09 authors/year), Dermatology (0.10 authors/year), Orthopedic Surgery (0.10 authors/year), Plastic Surgery (0.10 authors/year), and Thoracic Surgery (0.10 authors/year). Graphical comparison of authorship growth, by specialty, is presented in [Figure 2](#fig2-2050312120915399){ref-type="fig"}, and by article type, within each specialty, is presented in [Figure 3](#fig3-2050312120915399){ref-type="fig"}.
{#fig2-2050312120915399}
{#fig3-2050312120915399}
Specialty factors associated with number of authors per article {#section12-2050312120915399}
---------------------------------------------------------------
Specialties with more graduates entering academic practice were associated with more authors per article. Each percent increase in graduates entering academics was associated to 0.11 (95% CI = \[0.04, 0.19\]) more authors per article (p = 0.01). Neither surgical specialty (p = 0.36) nor length of residency training (p = 0.11) was found to be associated factors in the adjusted model.
Discussion {#section13-2050312120915399}
==========
The goal of this study was to generate data that describe authorship trends in contemporary medical literature in order to open a dialogue about authorship practices. Ambiguous recommendations, and variability in institution-based customs and politics, make this a challenging topic that few have openly discussed and defined. Appropriate authorship designation is imperative for maintaining the credibility of medical research. However, since citation counting has become the established means to determine academic prominence, it has also become a system which can be "gamed" with appropriate awareness of the standards. The original metrics of academic productivity may now have ceased to hold the same relations to the outcome measures they were designed to assess.
In our evaluation of 31,877 case reports, 34,799 review articles, and 54,721 original research articles, we identified a global increase in the number of authors per article over the past 12 years across article types and specialties. Publication year was found to be an independent predictor of higher number of authors per publication. Each consecutive year is associated with 0.16 more authors per peer-reviewed article. Although it is conceivable that increased collaboration in big-data-driven studies is responsible for this change, certain findings appear to contradict that hypothesis. For example, the number of authors should not have increased for review articles, as the complexity or collaborative efforts needed to write this type of publication have not increased over the last decade. This leads us to suspect that authorship inflation may have contributed to the observed increased authorship trend.
We were also interested to know how authorship trends compared between surgical and non-surgical specialties, which specialty characteristics impact bibliometric measures, and the nature of this impact. While all physicians first undergo similar 4-year training during medical school, training lengths and academic interests diverge dramatically during residency training. Certain specialties and specific training programs place greater emphasis on research and evaluate for participation when screening applicants.^[@bibr19-2050312120915399]^ Therefore, we suspected more competitive residency specialties and those known to place more emphasis upon academic participation would have more authors per article and/or show a greater number or growth in authorship.
The only characteristic we observed to be an associated factor was percentage of graduates with academic involvement after residency. Physicians who spend greater time in academic environments are more likely to be influenced by, and participate in, interdisciplinary medical research based upon their surroundings. Interestingly, surgical specialties, or the perceived competitiveness of a specialty's match, do not appear to be a deterministic factor of authorship quantity. The five specialties with the greatest number of authors per article (adjusted for article type and year of publication) range in competitiveness, procedural capacity, and primary care designations---Neurology, Radiation Oncology, Pathology, Psychiatry, and Internal Medicine. Both environmental influences during training and self-selection of residency applicants likely influenced specialty-publishing behaviors.
Measuring productivity in academics {#section14-2050312120915399}
-----------------------------------
The primary way an individual researcher's productivity, regardless of academic discipline, is quantified by way of the H-index. The H-index is an author-level metric derived from citations of his or her published works.^[@bibr20-2050312120915399]^ Authors whose names appear on more papers have a greater likelihood of reaching a higher H-index with more citations, which in turn benefits them in both their academic standing and potentially their ability to secure research time and funding. The H-index is not immune to inflation by means of self-citations, so authors have a theoretical incentive to cite their previous work. Although our current study did not aim to address the issue of self-citations, we did observe a positive, independent association between increased authorship and number of citations an article received. There is inherent bias to the simple H-index where self-citations can artificially increase an author's H-index, and hence modifications excluding self-citations have been proposed.^[@bibr21-2050312120915399]^
Another strategy to increase one's H-index is to be an author on a piece of work with a broader audience with potential for greater visibility. It was notable that, overall, and for most specialties, case reports initially had more authors than review articles, and with time, this trend reversed---with review articles now having more authors than case reports. We suspect this shift may be related to the greater awareness that review articles garner more citations (mean RCR for review articles = 2.98, as compared with that of case reports = 0.54). Though each measure of publication impact is prone to being manipulated by authors seeking increased visibility and recognition, all are largely based on the reputation of the authors in the byline and the publications in which they are published.^[@bibr22-2050312120915399]^ There has also been a significant movement toward quantitative metrics including Altmetric for measure of publication prolificacy and post-publication value. This metric takes into account the impact of electronic citations in the ever-increasing era of electronic communication and social media prominence.^[@bibr23-2050312120915399]^
Exploring the "publish or perish" mentality {#section15-2050312120915399}
-------------------------------------------
The professional stature of an academician is measured by the number of publications, impact factor of the journals publishing the articles, and by number of citations the work receives following publication. It is these mounting pressures to raise or maintain academic reputation that can lead to authorship inflation.^[@bibr24-2050312120915399]^ Honorary or gift authorships---the inclusion of a well-respected researcher for political reasons or to bring more merit to an article---and inclusion of junior authors for purposes of building academic rapport are both known forms of authorship inflation.
In our current system of calculating an author's H-index, all co-authors are rewarded the same credit regardless of the number and order in which these authors' names appear. Authorship inflation and self-citation undoubtedly allow researchers to gain greater visibility within their field, and individuals may feel pressured to partake in this practice because those not participating are at a relative disadvantage compared to those who do. There is little incentive to be conservative with authorship when the alternative is more likely to yield greater benefit in visibility and subsequent citation potential.
The ICMJE released guidelines on required authorship contributions, the most well-established set of guidelines used across most medical journals, was published in an effort to standardize practice and guide authors faced with authorship disputes. Despite the ICMJE's attempt to regulate authorship assignment, there is little evidence to show ICMJE guidelines have altered any publication practices.^[@bibr25-2050312120915399]^ Not all medical journals require compliance with ICMJE Uniform Recommendations, and authors submitting to journals requiring compliance may not fully understand the specified recommendations.^[@bibr25-2050312120915399],[@bibr26-2050312120915399]^ For example, a survey of over 300 healthcare professionals found that 33% of respondents believe "general supervision of a research lab" was sufficient to merit authorship.^[@bibr27-2050312120915399]^ Interestingly, there was no difference in the level of understanding of appropriate authorship criteria when respondents were aware or unaware of the guidelines. In a separate survey of 119 American medical, veterinary, and dental students enrolled in research fellowships at either the NIH or a sponsored academic medical center, 66% of respondents reported never receiving formal training in authorship guidelines.^[@bibr28-2050312120915399]^
Although there is a need for greater ethical publishing practices, simple promotion of good practices is not enough and unlikely to result in significant changes on its own. Dedicated training, formal discussion with research mentors, and regarding authorship may be valuable to both seasoned researchers and trainees. In order to develop a generation of ethical clinician scientists, current leaders must lead by example and realize their actions have a trickle-down effect on those they train. Physicians and scientists who understand are more likely to attract top personal to their team. When they set the right example for their team, they help foster an environment that encourages good citizenship, and motivates better performance and greater innovation.
Study limitations {#section16-2050312120915399}
-----------------
Our study has several recognized limitations. We did not incorporate specific measures of research complexity such as number of collaborating institutions and/or departments, or number of research participants per study by adding those noted as contributors in the acknowledgment sections. As such, we can only speculate whether authorship inflation occurred in original research articles independent of increased research complexity, as this inflation process is likely the reason for the increase seen in the review articles and case reports. Nevertheless, previous studies limited to specific research focuses have found evidence implying inflation as a known culprit of the growing authorship trend seen.^[@bibr12-2050312120915399]^ Another obstacle we encountered in our study was designing an analysis where the specialty journals were equally "impactful." Journal impact factor is reflective of the annual mean citations of articles published. Therefore, undertaking analysis involving only articles from "top" impact journals lends a bias to the analysis toward journals which publish more review articles and journals which are broader in scope. Despite the strict criteria for limiting journals, the citation values between specialty journals likely still vary. Journals were evaluated individually for inclusion or exclusion in our study based on content, specialty relevance, and impact factor as a surrogate for reach. Finally, article-type classifications were determined by indexed classifications in the PubMed database. However, because we evaluated articles over a 12-year period and included a diverse set of journals for analysis---article-type classification discrepancies---we have theoretically affected all specialties and journals equally, hence minimizing this potential bias in the comparative results.
Conclusion {#section17-2050312120915399}
==========
There has been substantial growth in the authorship bylines of contemporary medical literature, much of which cannot be fully explained by increasing research complexity alone. The increasing authorship trend is a reflection of the contemporary research landscape---one with greater complexity and collaboration and also one with more competition from lessening resources. While most are familiar with unethical publication practices such as plagiarism, falsification, and non-disclosure, less are aware of the realm of appropriate authorship. Trainees to seasoned scientists participating in medical research have likely faced or will face questions in this area. As the research environment continues to evolve with advanced tools and increased academic pressures for publication, it may be prudent to investigate how academic productivity is measured and rewarded based on contributions to quality medical literature.
Special thanks to the National Institutes of Health Office of Portfolio Analysis for providing early access to their portfolio analysis platform, guiding us through initial data queries and for their review of the manuscript. Finally, we would like to acknowledge Rachael Billock, PhD candidate, for her review of the statistics, and Daniel Marchalik, MD, for his edits and comments.
**Declaration of conflicting interests:** The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
**Funding:** The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Intramural Research Program of the National Institutes of Health (NIH) and the NIH Medical Research Scholars Program, a public--private partnership supported jointly by the NIH and generous contributions to the Foundation for the NIH from the Doris Duke Charitable Foundation, The American Association for Dental Research, the Colgate-Palmolive Company, Genentech and alumni of student research programs, and other individual supporters through contributions to the Foundation for the National Institutes of Health.
**ORCID iD:** Julie Y An  <https://orcid.org/0000-0001-8389-4908>
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
The incidence of melanoma in Europe has been steadily increasing, more rapidly than for any other cancer, during recent decades. Incidence of melanoma deeply varies across Europe, with the highest incidence in Scandinavian countries.
Lifetime risk of developing melanoma within the entire caucasian population is estimated to be 1:535 individuals. Incidence is higher in individuals with fair skin who have been exposed to high levels of UV-B radiation. Mortality due to metastatic melanoma has increased rapidly in both males and females; such a disease accounts for only 4% of skin cancers, but for 79% of skin-cancer related deaths. In particular, mortality rates are higher in males than in females; mortality has doubled in males over the past 25 years. Again, death rates are higher in individuals with fair skin. From 2002 to 2006, the median age at death due to metastatic melanoma of the skin was 68 years. Both incidence and mortality rates are highest in Australia and New Zealand. The 5-year survival rate for patients with advanced melanoma is less than 10% \[[@B1]-[@B6]\].
In melanoma patients, survival is strongly related to tumor thickness; earlier diagnosis and complete excision of lesions may thus lead to a decline in mortality for such a disease. The most effective tool in fighting malignant melanoma is detecting the disease with a depth of ≤ 1 mm (Breslow thickness) and without ulceration, which is associated with an excellent prognosis (95% 5-year survival rate), in comparison to detection of a malignant melanoma with a depth of \> 4 mm (40% 5-year survival rate), as reported by the American Joint Committee on Cancer \[[@B7]\].
Diagnostic accuracy for pigmented skin lesions using the naked eye has been estimated to be about 60% \[[@B8]\]. To improve the accuracy of melanoma diagnosis, a variety of diagnostic instruments has been developed: dermoscopy, multispectral imaging, confocal laser microscopy, ultrasonography, optical coherence tomography, or magnetic resonance imaging \[[@B9]\].
From its introduction in 1990s, dermoscopy or epiluminescence microscopy (ELM), a non-invasive tool for cutaneous screenings, has become the most popular technique for early diagnosis of melanoma \[[@B10]-[@B33]\], also reducing the excess of cases addressed to excision biopsy \[[@B34]\]. However, ELM is an useful tool in expert hands. As stated in a systematic review by Kittler and colleagues, dermoscopy does improve the diagnostic accuracy for melanoma in comparison with eye-based inspection, when used by experienced examiners only \[[@B35]\]. Recentely, a meta-analysis evaluating diagnostic accuracy for melanoma confirmed that sensitivity was much higher for dermoscopy (0.90; 95% confidence interval \[CI\], 0.80-0.95) than for naked eye examination alone (0.71; 95% CI, 0.59-0.82), with an estimated difference of about 0.18 (95% CI, 0.09-0.27; P = 0.002) \[[@B36]\]. Conversely, there was no statistical evidence of a significant difference in specificity \[dermoscopy 0.90 (95% CI, 0.57-0.98) *versus*naked eye examination 0.81 (95% CI 0.48-0.95); difference 0.09 (95% CI, 0.06-0.23, P = 0.18)\] \[[@B36]\]. In addition, a meta-analysis of 30 studies using dermoscopy showed a sensitivity of 0.88 (95% CI, 0.87-0.89) and a specificity of 0.86 (95% CI, 0.85-0.86) \[[@B37]\].
Since sensitivity and specificity of dermoscopy significantly vary among different studies without reaching the 100% of validity, additional non-invasive techniques, such as spectrophotometric intracutaneous analysis (SIA) and spectrophotometry (SPT), have been introduced to improve early diagnosis of melanoma.
The SIA technique (SIAscopy, as named by Moncrieff et al. \[[@B38]\]), which was developed using the model of light transport \[[@B39],[@B40]\], is aimed at differentiating melanomas from non-melanoma skin cancers, benign melanocytic naevi and other pigmented lesions such as seborrhoeic keratosis \[[@B38]\].
The spectrophotometry is a method that evaluate the color of a lesion by measuring its reflectance as a function of the wavelength. A pioneer in this field was Marshall \[[@B41]\] who described a comprehensive analysis of pigmented lesions under both ultraviolet and infrared radiations. In such a study, he stated that the infrared photographs tends to discriminate the different types of lesions, with melanoma showing a relatively high degree of correlation with low infrared reflectance.
The SPT (Spectroshade, MHT Verona, Italy) system allows to extract information regarding the skin microarchitecture, like the distribution, position, and quantity of blood, collagen and melanin throughout the epidermidis and papillary dermis, converting data from reflected near infrared radiation into *in vivo*histological images of the lesion \[[@B38]\]. The SPT system provides information including a series of 15 multispectral images into the near infrared bandwidth. On this regard, permanence of the image into the infrared area of the spectrum has been demonstrated to represent a significant indication of malignancy of the pigmented lesion under examination. Seven parameters (mean reflectans, MR; variegation, V; area, A; dark area ratio, DAR; dark island reflectance, DA; dark distribution factor, DDF) are processed automatically by software, describing the pigmented lesion features. Three spectral areas play a major role in quantification of all these parameters: 584 nm, where the highest contrast between lesion and the surrounding skin is determined; 650-750 nm, where a pigmented lesion can be better discriminated since the light adsorption due to melanine is much higher than that due to blood; 750-950 nm, where the lesion color is quantified in the near infrared, a spectral area invisible to the clinician\'s eye. After exposing the skin to visible light and infrared radiation, the spectrophotometric system converts the reflected radiations into *in vivo*histological images of the lesion by means of a sophisticated computer algorithm. Such an image allow to extract information regarding the microarchitecture of the cutaneous alteration under examination.
Starting from our experience in the use of both ELM and SPT, we intend to define the role of spectophotometric diagnosis in the early detection of cutaneous melanoma.
Methods
=======
In recent years, physicians of the Melanoma Cooperative Group at the National Cancer Institute of Naples have performed a considerable amount of visits for cutaneous screening (about 10.000 visits per year). For the evaluation of clinically-relevant pigmented lesions, epiluminescence microscopy (ELM) was carried out using a Molemax II (Derma Instruments, Vienna) video-dermatoscope.
According to the criteria provided by our group for the treatment of cutaneous pigmented lesions \[[@B42]\], each patient was interviewed in order to obtain a personal profile (date of birthday, place of residence, type of work, history of sun exposure and sun burning, familial recurrence of skin lesions), before undergoing a clinical total body skin examination. Cutaneous pigmented lesions were classified according to the macroscopic ABCDE criteria; when at least two of the ABCDE criteria occurred, individuals were addressed to ELM-based dermoscopy, in order to generate a risk classification. Such an evaluation was made by expert dermatologists (at least 3 years of experience). Lesions were classified as non-melanocytic lesions (such as angiokeratoma, verrucous naevus, pigmented basal cell carcinoma, seborrhoeic keratosis, angioma, keratoacantoma, solar keratosis, etc.) or melanocytic lesions \[such as compound naevus, intradermal naevus, papillomatous compound naevus, Spitz naevus, blue naevus (without pigment network); junctional naevus, lentigo simplex, pigmented spindle cell naevus of Reed, naevus spilus, cockarde naevus, atypical naevus, malignant melanoma (with pigment network)\] (Table [1](#T1){ref-type="table"}).
######
ELM-based classification of cutaneous lesions and corresponding histology
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[CUTANEOUS LESIONS]{.smallcaps} [HISTOLOGICAL LESIONS]{.smallcaps}
--------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------
**Non-melanocytic lesions** angiokeratoma, verrucous nevus, pigmented basal cell carcinoma, seborrhoeic keratosis, angioma, keratoacantoma, solar keratosis, etc.
**Melanocytic lesions**\ compound nevus, intradermal nevus, papillomatous compound nevus, Spitz nevus, blue nevus.
**(without pigment network)**
**(with pigment network)** junctional nevus, lentigo simplex, pigmented spindle cell nevus (Reed nevus), nevus spilus, cockarde nevus, atipycal nevus, **melanoma**.
**Other lesions** squamous carcinoma, trichoepithelioma, Becker\'s nevus, Ota\'s nevus, Ito\'s nevus, Halo nevus.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Melanocytic lesions were further classified as very low, low, medium, high, and very high risk lesions \[[@B18]\]. Characterization and classification of each lesion were based on the presence or absence of typical ELM features (Table [2](#T2){ref-type="table"}) \[[@B18]\]. According to our working formulation \[[@B42]\], each category was further classified as A (not clinically suspicious) or B (clinical suspicious), on the basis of the clinical history of the lesion (a suspicious history is defined when a lesion presents a modification of its clinical parameters during the past 6 months).
######
ELM-based risk classification for cutaneous pigmented lesions
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[MELANOCYTIC LESIONS]{.smallcaps} [ELM FEATURES]{.smallcaps}
----------------------------------- -----------------------------------------------------------------------------------------------------------------------------------------------------------------
**Type 1**\ Lesion with a pigment network and any of the classical ELM features specific for melanoma (pseudopods, radial streaming, blue-gray veil, atypical vessel, etc.)
**(very high risk)**
**Type 2**\ Lesion with a pigment network and subtle new ELM features that may suggest melanoma but often are also seen in atypical nevi.
**(high risk)**
**Type 3**\ Lesion with a pigment network carrying subtle perturbations that can be detected in atypical naevus as well as in melanocytic hyperplasia.
**(medium risk)**
**Type 4**\ Lesion with a benign appearing network.
**(low risk)**
**Type 5**\ Lesion with a benign appearing network and with a globular pattern or another benign ELM pattern.
**(very low risk)**
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Surgical excision was advised for high and very high risk lesions. In type A of medium, low, and very low risk lesions, surgery may be justified on the basis of cosmetic and/or functional reasons (in absence of contraindications for surgical excisions). Anyway, patients with medium risk lesions entered a close follow-up program. Surgery was always recommended for all type B lesions \[[@B42]\]. In case of advise for surgery of a melanocytic lesion, patients were invited to participate in our study; a written informed consent was obtained before patient\'s enrollment into the study.
Fifty-four individuals with cutaneous pigmented lesions, who were addressed to surgery after ELM classification, were also evaluated by spectrophotometry (SPT), using the SpectroShade system (MHT, Verona, Italy). Comparisons between SPT and ELM results were conducted in order to verify whether SPT may highlight, through multispectral image analysis, new features of pigmented lesions (content and distribution of the adsorbing components of the skin, such as melanin and haemoglobin), which might be useful to improve the early diagnosis of melanoma. Table [3](#T3){ref-type="table"} summarizes the characteristics of the SPT classes and presents the correlation with the ELM risk classification.
######
SPT diagnostic classes and corresponding ELM risk classes
[SPT CLASSES]{.smallcaps} [PROBABILITY OF MELANOMA]{.smallcaps} [CORRESPONDING ELM RISK CLASS]{.smallcaps}
--------------------------- --------------------------------------- --------------------------------------------
**1** no melanoma Non Melanocytic and Type 5
**2** doubtful melanoma (\< 25%) Type 4
**3** suspected melanoma (\< 75%) Types 3 and 2
**4** probable melanoma (\> 75%) Type 1
The study was reviewed and approved by the Ethycal Committee at the National Cancer Institute of Naples (Italy).
Statistical methods
-------------------
SPT sensitivity and specificity as well as positive and negative predictive values were determined according to a computer-assisted statistical analysis using the SPSS software for Windows, 8.0 version (SPSS Inc., Chicago, USA). Taking histology as standard, a lesion was considered true positive (TP) or true negative (TN) if results from SPT or ELM and histology were consistent (in other words, a concordance was achieved using such two diagnostic approaches). Conversely, a lesion was considered as false positive (FP) or false negative (FN) if histology did not confirmed the classification of SPT/ELM (in other words, SPT/ELM was positive but histology was negative or SPT/ELM was negative but histology was positive). Sensitivity was calculated as TP/TP+FN; specificity was calculated as TN/TN+FP; positive predictive value was calculated as TP/TP+FP; negative predictive value was calculated as TN/TN+FN.
Results
=======
In a period of one year, total-body skin examinations for detection of early malignant melanoma were carried out at National Cancer Institute of Naples; they allowed to identify cutaneous pigmented lesions to be addressed to surgical excision and histological examination in 54 patients \[35 females and 19 males; median age, 41 (range, 19-73 years)\]. Before surgery, lesions were evaluated by both ELM and SPT approaches.
According to ELM criteria \[[@B42]\], all 54 cutaneous pigmented lesions were classified as melanocytic lesions; according to the risk criteria (Table [3](#T3){ref-type="table"}), they were classified as very low risk (N = 11; 20%), low risk (7; 13%), medium risk (1; 2%), high risk (21; 39%), and very high risk (14; 26%) lesions. The comparison between ELM-based classifications and histological results among the 54 excised lesions gave an agreement of 100% for very low risk lesions (6 papillomatosus compound naevus and 5 intradermal naevus), 100% for low risk lesions (7 junctional naevus or compound naevus), 100% for medium risk lesion (1 lentigo simplex), 90.4% for high risk lesions (19 atypical naevus and 2 Reed naevus), and 85.7% for very high risk lesions (12 melanoma and 2 atypical naevus). Overall, agreement between histology and ELM diagnosis was 92.5% (Table [4](#T4){ref-type="table"}).
######
Agreement between ELM diagnosis and histology
Type of lesion ELM Diagnosis Histological Diagnosis Positive/Negative Agreement %
------------------------- --------------- ------------------------------------------ -------------
Non-melanocytic lesions \- -/- \-
Melanocytic lesions 54 50/4 92.5
Very low risk 11 11/- 100
Low risk 7 7/- 100
Medium risk 1 1/- 100
High risk 21 19/2 90.4
Very high risk 14 12/2 85.7
TOTAL 54 50/4 92.5
Before surgery, the same 54 selected pigmented lesions were evaluated by SpectroShade software, which automatically classified them in four different classes: group 1, non melanoma; group 2, doubtful melanoma (\< 25%); group 3, suspected melanoma (\< 75%); group 4, probable melanoma (\> 75%). Since criteria for classification into such groups appeared to be not particularly accurate, we decided to modify such a classification and correlate it to the ELM-based risk assessment of each specific cutaneous lesion. Therefore, we defined the following classes: group 1, non melanocytic and very low risk lesions; group 2, low risk lesions; group 3, medium and high risk lesions; group 4, very high risk lesions (Table [3](#T3){ref-type="table"}).
Again, the comparison between SPT classifications and histological results among the 54 excised lesions gave an agreement of 66.6% (3 lesions) in group 1, 100% (5 lesions) in group 2, 57.1% (28 lesions) in group 3, and 44.4% (18 lesions) in group 4 (Table [5](#T5){ref-type="table"}). Among all pigmented lesions, agreement between histology and SPT diagnosis was low (57.4%) (Table [5](#T5){ref-type="table"}). Consistently, the agreement between ELM and SPT classifications was also quite low (overall, 68.5%) (Table [6](#T6){ref-type="table"}).
######
Agreement between SPT diagnosis and histology
Groups SPT Diagnosis Histological Diagnosis Positive/Negative Agreement %
--------- --------------- ------------------------------------------ -------------
Group 1 3 2/1 66.6
Group 2 5 5/0 100
Group 3 28 16/12 57.1
Group 4 18 8/10 44.4
TOTAL 54 31/23 57.4
######
Agreement between SPT and ELM classifications
Groups SPT Diagnosis ELM Diagnosis Positive/Negative Agreement %
--------- --------------- --------------------------------- -------------
Group 1 3 2/1 66.6
Group 2 5 5/0 100
Group 3 28 16/12 57.1
Group 4 18 14/4 77.7
TOTAL 54 37/17 68.5
As summarized in Table [7](#T7){ref-type="table"}, SPT-based classifications and histological results were evaluated for both sensitivity and specificity of this method among the 54 excised lesions. The sensitivity and specificity of SPT in detecting melanoma (group 4) were 66.6% and 76.2%, respectively; moreover, the positive and negative predict values were 44.4% and 88.9%, respectively (Table [7](#T7){ref-type="table"}). Since the rate of SPT diagnoses was progressively higher as the thickness of the lesions increased (Table [8](#T8){ref-type="table"}), the sensitivity and specificity of this method in detecting high risk lesions (atypical naevi and melanomas) remained lower than expected.
######
Sensitivity and specificity for SPT in group 4 lesions
Histology positive Histology negative
-------------- -------------------- -------------------- --------------------
SPT positive 8 10 Sensitivity = 66.6
SPT negative 4 32 Specificity = 76.2
######
Breslow thickness of melanomas diagnosed in the series of 54 cutaneous lesions
[BRESLOW THICKNESS CATEGORIES]{.smallcaps} [N. OF CASES]{.smallcaps} [MEDIAN BRESLOW THICKNESS]{.smallcaps} [RANGE]{.smallcaps}
-------------------------------------------- --------------------------- ---------------------------------------- ---------------------
**\< 1 mm** **5** **0.7** **0.3-1.0**
*1* *1.0* *-*
**1.01 to 2 mm** **5** **1.45** **1.2-2.0**
*5* *1.45* *1.2-2.0*
**2.01 to 4.0 mm** **2** **2.25** **2.1-2.4**
*2* *2.25* *2.1-2.4*
**\> 4.01 mm** **0** **0** **0**
**TOTAL** **12** **1.3** **0.3-2.4**
The 8 cases diagnosed by SPT are indicated in *italic*.
Discussion
==========
Classification of pigmented lesions is actually based on both clinical evaluation, by naked-eye, and dermoscopy of the skin features; however, diagnosis strongly depends on the experience of physicians. At present, the ELM methodology, even in the hands of experienced examiners, does not have 100% sensitivity and specificity (overall, 88% sensitivity and 86% specificity \[[@B37]\]). To improve the early detection of melanoma, new diagnostic tools providing an automated classification of pigmented skin lesions (in other words, new devices which could be also used by non-expert physicians) have been proposed. Among others, the SPT is considered as the most promising method for such a purpose.
Although the series of analyzed patients in the present study was quite small, our findings seem to indicate that the SPT may have a limited role into the early diagnosis of melanoma (66.6% sensitivity; 76.2% specificity; 44.4% positive predicitive value; 88.9% negative predictive value). About sensitivity, one melanoma (0.8 mm breslow thickness, with a large dermal regression) was classified in group 1 (very low risk) by SPT; the other three melanomas (median breslow thickness, 0,4 mm; range 0,3-0,7 mm) were classified in group 2 by SPT. Similarly, specificity was quite low since only 8 out of 18 lesions, which have been classified in group 4 by SPT, were really malignant after histological diagnosis (consequently, the agreement with histology was also low).
Another important issue is the difficulty to perform the SPT evaluation on palpable lesions such as papillomatous naevi. In fact, 6 of these latter lesions and 3 dermal naevi were diagnosed in group 3 by SPT; therefore, this contributed to lower the agreement with the histology also in such a group.
Considering previous studies using the SPT approach, some difference was observed among the image acquisition systems of the various instruments. However, the Spectroshade^®^(MHT, Verona, Italy), which was used for screening in our series, and the SIAscope (Astron Clinica, Cambridge, UK), which has been commonly used worldwide, are quite similar; both spectrophotometers are able to acquire images at wavelengths between 483-950 nm (Spectroshade) \[[@B43]\] or 400-1000 (SIAscope) \[[@B38]\] and both of them use computer algorithms to analyze quite similar parameters of the pigmented lesions. Therefore, it is possible to compare our results with those reported in literature, using these two spectrophotometric systems. Table [9](#T9){ref-type="table"} summarizes the results from the main studies published in past years \[[@B38],[@B44]-[@B50]\]~.~. Farina et al. (2000) \[[@B45]\] reported the first large study (including 237 patients) describing a pretty good sensitivity (80%) but a low specificity (46%). Better results were reported by Elbaum et al. (2001) \[[@B47]\] in a series of 246 patients; in this case, the authors used the MelaFind^®^(Electro-Optical Sciences Inc., Irvington, New York, USA), a multispectral digital dermoscope with a function similar to the others spectrophotometers, which was described to achieve a sensitivity of 100% and a specificity of 85%. Analogously, high values of sensitivity and specificity were presented by two others studies (Moncrieff et al., 2002 \[[@B38]\]; Govindan et al., 2007 \[[@B49]\]); however, Moncrieff and colleagues concluded that the SIAscope did not add accuracy in melanoma diagnosis in comparison with the dermatologist\'s long experience.
######
Summary of the most important studies on sensitivity and specificity of SPT
Reference N. Patients Sensitivity Specificity
------------------------------------- ------------- ------------- -------------
Tomatis S et al., 1998 \[[@B44]\] **44** **89%** **88%**
Farina B et al., 2000 \[[@B45]\] **237** **80%** **46%**
Wallace VP et al., 2000 \[[@B46]\] **47** **84%** **91%**
Elbaum M et al., 2001 \[[@B47]\] **246** **100%** **85%**
Moncrieff M et al., 2002 \[[@B38]\] **311** **83%** **80%**
Haniffa MA et al., 2007 \[[@B48]\] **860** **87%** **91%**
Govindan K et al., 2007 \[[@B49]\] **886** **94%** **80%**
Glud M et al., 2009 \[[@B50]\] **65** **100%** **59%**
The present study **54** **66%** **76%**
Consistently with this latter indication, Haniffa et al. (2007) \[[@B48]\] reported high values of sensitivity (87%) and specificity (91%) for SPT among 860 lesions examined using the SIAscope spectrophotometer; when the same lesions were however evaluated by a dermatologist with a dermatoscope and at least 3 years of experience, sensitivity and specificity resulted to be even higher (94% and 91%, respectively). Since diagnostic accuracy of dermoscopy and SPT was similar, again authors suggested that a spectrophotometer does not improve the diagnostic ability of experienced dermatologists \[[@B48]\]. Finally, the recent study by Glud et al. (2009) \[[@B50]\] revealed a 100% sensitivity but a 59% specificity for SPT (as in Farina et al., 2000 \[[@B45]\]), further indicating that such an approach may overestimate the proportion of possible malignant lesions (high amount of false positives).
Although the SPT is still considered as a valuable diagnostic tool, the low specificity of such a methodology (probably, due to the interference of seborrhoeic keratoses, which are not always recognized as benign lesions) represents the main hamper for the introduction of such a diagnostic tool into the clinical practice. No seborrhoeic keratosis was instead present in our series; therefore, we may speculate that the difference in diagnostic accuracy between ELM and SPT was due to additional factors in our study. Probably, the automated software of the SPT might use some parameters that do not discriminate the features of melanoma or overestimate the role of others skin components (blood, melanin, and collagen in the infrared band) as markers of malignancy. Finally, the agreement between the two screening methods was unsatisfactory (68.5%), and the high number of false positives in group 3 and 4 makes the ELM evaluation as the best technique to help clinicians in early diagnosis of cutaneous melanoma. Consistently, SPT should be utilized in clinical trials only.
Conclusions
===========
The spectrophotometry (SPT) allows to accurately define the skin microarchitecture features. However, the low accuracy of SPT for either thin melanocytic lesions (particularly, those with regression) or palpable lesions (such as papillomatous and dermal naevi) clearly indicates that dermoscopy remains the best diagnostic tool for the preoperative diagnosis of pigmented skin lesions.
Competing interests
===================
PAA participated to advisory board of the Bristol Myers Squibb and GSK; he received honoraria from Schering Plough. The remaining authors declare that they have no competing interests.
Authors\' contributions
=======================
PAA: Conception and design, manuscript writing. MP: Data analysis and interpretation, provision of study material/patients. FA: Data analysis and interpretation, provision of study material/patients. IDM: Collection and assembly of data. CC: Provision of patients, collection and assembly of data. AD: Provision of patients. ES: Provision of patients. SM: Provision of patients. MDG: Provision of patients. AV: Provision of patients. GP: Manuscript writing. NM: Conception and design, final approval of manuscript. All authors were involved in manuscript writing and provided final approval of the manuscript.
Pre-publication history
=======================
The pre-publication history for this paper can be accessed here:
<http://www.biomedcentral.com/1471-5945/10/5/prepub>
Acknowledgements
================
Authors would like to thank MHT (Medical High Technologies S.p.A., Verona, Italy) for providing technical support in using the Spectroshade. A special thank to Alessandra Trocino, for providing excellent bibliography service and assistance, and Ilenia Visconti, for data management. Work was supported by Italian Ministry of Health \"Progetto Ricerca Finalizzata\" and Associazione UMANA Onlus.
| {
"pile_set_name": "PubMed Central"
} |
Aims and Scope
==============
*Clinical Case Reports* is different to other case report journals. Our aim is to directly improve global health outcomes and share clinical knowledge using case reports to convey important best practice messages. We welcome case reports, clinical images, and procedural videos from all areas of Medicine, Nursing, Dentistry, and Veterinary Science.
*Clinical Case Reports* aims to publish cases which illustrate the use of important systematic reviews. *Clinical Case Reports* also has a close relationship with the Cochrane Collaboration and case report submissions which relate to Cochrane systematic reviews are a priority for the Journal. If your submission has an important clinical message we would like to read it!
We focus on common and unusual clinical situations, and encourage critical refl ection and critique of practice, and have a particular interest in reports that draw on the use of clinical guidelines. Why are we interested in common as well as uncommon clinical scenarios? Our view is that by publishing important common clinical scenarios managed according to agreed best practice principles we will help inform a broader clinical audience than the more esoteric reports.
When writing your case report, please consider that our mission "...is to contribute to quality patient care and health professional education through the dissemination of clinical practice scenarios. We want to support clinicians in providing patients with the best possible care, and ultimately we aim to provide a reservoir of case reports that can also be used in teaching and for secondary research purposes". For further information, please refer to our Author Guidelines available at <http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2050-0904/homepage/ForAuthors.html>
Open Access and Copyright
=========================
All *Clinical Case Reports* articles are fully open access: immediately freely available to read, download and share. If your paper is accepted, the author whom you identify as being the corresponding author will be presented with the option to sign an open access agreement (on behalf of all co-authors) to make the case report available under the terms of either the Creative Commons Attribution Non-Commercial (CC BY-NC) license or the Creative Commons Attribution-Non-Commercial- NoDerivs (CC-BY-NC-ND) license. RCUK or Wellcome trust funded authors will be directed to sign the Open Access Agreement under the terms of the Creative Commons Attribution (CC-BY) license in order to be funder compliant. For more information on the terms and conditions of these licenses please visit: [http://www.wileyopenaccess.com/details/content/12f25db4c87/Copyright--License.html](http://www.wileyopenaccess.com/details/content/12f25db4c87/Copyright–License.html)
Copyright on any research article published in *Clinical Case Reports* is retained by the author(s). Authors grant Wiley a license to publish the article and identify itself as the original publisher. Authors also grant any third party the right to use the article freely as long as its integrity is maintained and its original authors, citation details and publisher are identifi ed. The Open Access Agreement formalizes these and other terms and conditions of publishing research articles.
Disclaimer
==========
The Publisher and Editors cannot be held responsible for errors or any consequences arising from the use of information contained in this journal; the views and opinions expressed do not necessarily refl ect those of the Publisher and Editors, neither does the publication of advertisements constitute any endorsements by the Publisher and Editors of the products advertised.
Wiley Open Access articles posted to repositories or websites are without warranty from Wiley of any kind, either express or implied, including, but not limited to, warranties of merchantability, fi tness for a particular purpose, or non-infringement. To the fullest extent permitted by law Wiley disclaims all liability for any loss or damage arising out of, or in connection with, the use of or inability to use the content.
Editor-in-Chief
===============
Professor Debra Jackson
World Health Organization Collaborating Centre for Nursing, Midwifery & Health Development, Australia University of Technology, Sydney, Australia
Address correspondence to the Editorial Office: Email: <clinicalcases@wiley.com>
Strategic Advisory Board
========================
Professor Jo-anne Brien
St. Vincent\'s Hospital, University of Sydney, Australia
Dr Matthew R. Lewin
California Academy of Sciences, USA
Dr Karen Myrick
Quinnipiac University, USA
Dr Dan Silasi
Yale University, USA
Dr David Tovey
Cochrane Library, UK
Dr Ted Whittem
University of Melbourne, Australia
Dr Jen-Hung Yang
Tzu-Chi University, Taiwan
Associate Editors
=================
Syed Mukhtar-Un-Nisar Andrabi
Aligarh Muslim University, India
Arshad Anjum
Aligarh Muslim University, India
Prantar Chakrabarti
Medical College Kolkata, India
David Chitayat
University of Toronto, Canada
Elena Chiticariu
Victor Babeş University of Medicine and Pharmacy, Romania
Steven Crosby
MCPHS University, USA
Daniel Darbyshire
Salford Royal Hospital, UK
Alessandro Diana
Clinique Grangettes, Switzerland
Martine Docx
Queen Paola Children\'s Hospital Antwerp, Belgium
Marcello Donati
University of Catania, Italy
Ajit Goenka
Cleveland Clinic Foundation, USA
Khaled Mursi Hammoud
Cairo University, Egypt
Hans Carl Hasselbalch
Roskilde Hospital, University of Copenhagen, Denmark
Bosco M. A. Hon-ming
The Chinese University of Hong Kong, Hong Kong
Mahesh Iddawela
University of Melbourne, Australia
Kelli Innes
Monash University, Australia
Ricardo J. P. José
University College London Hospital, UK
Nikolaos Karidis
Freeman Hospital NHS, UK
Eduardo Jose Veras Lourenço
University of Rio de Janeiro State, Brazil
Addisu Mesfin
University of Rochester School of Medicine and Dentistry, USA
Ana Miguel
Centre for Research in Health Technologies and Information Systems, Portugal
Milija D. Mijajlovic
Clinical Center of Serbia and University of Belgrade, Serbia
Nils-Halvdan Morken
Haukeland University Hospital and University of Bergen, Norway
Gill Murphy
University of Western Sydney, Australia
Bichitra Nanda Patra
Postgraduate Institute of Medical Education and Research, India
Shane O\'Hanlon
University of Limerick, Ireland
Gregorio Patrizi
University of Rome, Italy
Fabio Pilato
Università Cattolica del Sacro Cuore, Italy
Vivek S. Radhakrishnan
Tata Medical Center, India
Arshad Rashid
MMIMSR, MM University, India
Leonardo Oliveira Reis
University of Campinas (UNICAMP), Brazil
Raffaele Serra
University Magna Graecia of Catanzaro, Italy
Jeffrey Silbiger
Mount Sinai School of Medicine, USA
John Soedirman
Admiraal de Ruyter Hospital, Netherlands
Alessandra Soriano
Campus Bio-Medico University of Rome, Italy
Vedran Stefanovic
Helsinki University Central Hospital, Finland
Tracey Straker
Montefi ore Medical Center, USA
Ivo Tremont-Lukats
University of Texas MD Anderson Cancer Center, USA
Prash Vas
Addenbrookes Hospital, UK
Martin Whyte
Royal Surrey County Hospital and University of Surrey, UK
Alex K. Wong
Keck School of Medicine of USC, USA
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Stillbirth and its consequences have attracted some research interest. However, there still remains an uncertainty particularly about how to advise medical staff of the necessary steps that should be taken around stillbirth bereavement care. Before 1970s, preventing bereaved parents from seeing their stillborn infant was standard practice. In a landmark paper, Lewis outlined advice for health practitioners on how to facilitate the grief process by allowing parents to create memories with the deceased baby, such as naming the baby, and particularly encouraging women to see their stillborn babies \[[@CR1]\]. It was suggested that seeing the dead baby was actually important for the grieving process. This approach was adopted as part of stillbirth bereavement care at the UK hospitals until the late nineties. However, a controversial study by Hughes et al. \[[@CR2]\], suggested that women who did not see or hold their child had a lower prevalence of depression in their subsequent pregnancy. These women also exhibited fewer symptoms of anxiety and post-traumatic stress disorder (PTSD). The significance of this study was challenged by Brabin \[[@CR3]\] on the grounds of inconclusive statistical differences and validity issues. Despite this, and without further empirical evidence, NICE (2007) guidance changed and the informed practice was not to encourage mothers to see or hold their stillborn infant \[[@CR4]\]. Voluntary organisations (e.g. SANDS, UK) campaigned for the parents' right to be offered to see or hold their baby, brought about change in the guidance, and subsequently, informed practice was to offer the options to parents to choose. These polarised views in dealing with stillborn care particularly for stillbirth bereavement care are still evident, however, in research informed guidance. Recently, in a review of the stillbirth and its management, it was suggested that parents need to be supported around approaching their stillborn baby \[[@CR5]\]. Other recent findings found that, mothers who saw and held their stillborn babies, experienced lower levels of anxiety and depressive symptoms \[[@CR6]\]. However during pregnancy subsequent to stillbirth, mothers appeared to have less depressive symptomology but more symptoms of anxiety if they had seen or held their stillborn babies. Similarly, an overall positive effect of having held a stillborn baby was found, particularly for births after 37 weeks gestation \[[@CR7]\]. A recent study also found that mothers felt more comfortable and less frightened if the health care staff supported assumptive bonding by simply offering the baby to the mother, without emphasising the choice of whether she wanted to see or hold the baby \[[@CR8]\]. This may be because the act of choosing implies that the mothers are doing something that they are not supposed to do. Furthermore Ryninks and colleagues \[[@CR9]\] also highlighted that although contact with the stillborn infant caused distress to some women, they acknowledged the importance of such contact and proposed that this was the right decision to take. However, we still need to understand the individual circumstances surrounding the difficult decisions around stillbirth and use this to inform clinical guidance so that consistent reliable care is given to women.
The majority of studies used to inform stillbirth bereavement care are quantitative in nature. Recent qualitative studies have provided a closer look into the woman's own experiences in relation to the care and support that she received from health practitioners \[[@CR10]\], and the mother's view on their interactions with the health care staff before and after stillbirth \[[@CR11]\], as well as with their parenting experiences \[[@CR12]\]. However, to date no other research with an existential focus has looked at what stillbirth actually means to women. This is an important aspect that is missing from the research literature, considering individual differences dealing with trauma \[[@CR13]\]. Accounts of individuals will offer a useful insight into the meaning of stillbirth experience and the hard decisions that mothers have to make in that context.
Furthermore current research also suggests that mothers become anxious in their subsequent pregnancies and that this anxiety may have prolonged consequences for both mothers and the subsequent infant, for example increased psychopathology \[[@CR14], [@CR15]\] and disorganised infant attachment in subsequent infants \[[@CR16]\].
Few studies have looked at the subsequent pregnancy and the ensuing relationship with the infant. One follow-up study drew attention to the mothers' different perceptions and their attitudes toward the infants born subsequent to a stillbirth in comparison to their other children (e.g. 'the vulnerable child' or 'the replacement child') and called for further qualitative studies where a mother's view of her subsequent child, and individual differences between mothers, could be understood and identified during stillbirth bereavement care \[[@CR17]\]. Similarly, another recent qualitative study has also emphasized the impact of a loss of an infant on the subsequent parenting of mothers \[[@CR18]\], and indicated a paradoxical pattern (trying to hold their subsequent child emotionally close, but aloof at the same time) in their parenting styles. However, this emerging area of research requires further investment in order to understand the factors that may be affecting a mother's relationship with the subsequent infant, including an examination of parenting experiences.
Therefore, this study will focus on the meaning of the stillbirth experience to women and its influence on the subsequent pregnancy and subsequent parenting from the mothers' own perspectives. The primary questions of this study are: How do women make sense of their experiences of stillbirth? How do mothers who have previously experienced a stillbirth make sense of their relationship with a subsequent infant? The secondary question of the study is: how do the mothers' accounts of stillbirth and their relationship with subsequent children relate to the existential, cognitive (appraisal), and developmental (attachment) theories of psychology?
Reflexivity {#Sec2}
-----------
The first author experienced stillbirth 9 years ago and currently provides counselling services, particularly to women who have experienced prenatal and postnatal losses. She is a reliable Adult Attachment Interview, \[[@CR19]\] coder and aware that she has pre-formed ideas around the stillbirth experience. These were noted during the analysis of each participant's written accounts and during discussions which took place at supervision meetings with her supervisor. The second author is a phenomenological psychologist with an interest in the cultural context of personal experiences. The third author is a consultant clinical psychologist, interested in attachment issues. The fourth author is an applied developmental psychologist, whose specific interest lies in parenting and childhood.
Methods {#Sec3}
=======
Approach {#Sec4}
--------
A semi structured interview \[[@CR20]\] was chosen to gain a detailed picture of accounts of mothers' experiences. This interview was carried out by email to give participants time and space for remembering and reflecting on their own experiences without being overwhelmed. Written accounts were then analysed using interpretative phenomenological analysis (IPA) \[[@CR21]\]. IPA consists of case by case analysis of a number of accounts in a particular context and provides in-depth understanding of the meaning of a shared experience. For this reason IPA requires purposive sampling of a homogenous and small group \[[@CR21], [@CR22]\]. Analysis focused first on how women make sense of their experiences of their stillborn babies and their relationship with their subsequent infants. A second focus was on how women's account of their stillbirth experiences related to existential, cognitive (Appraisal) and developmental (Attachment Theory) components.
Recruitment {#Sec5}
-----------
After obtaining Ethical approval from the University of Birmingham for this study, the recruitment was carried out on internet-based social support websites -- Facebook - Stillbirth group, Twitter - After Stillbirth, The Birth Trauma Association (BTA) and a US based pregnancy loss support forum- Share.
Participants {#Sec6}
------------
A purposive sample of 6 women whose first pregnancy ended with a stillbirth and had since had a living infant (who was between the age of 4 months to 4 years old), took part in this study. The contextual details of the participants are presented in Table [1](#Tab1){ref-type="table"}. One of the participants had a twin pregnancy which ended with a stillbirth and premature birth, followed by neonatal death. However, like other participants this pregnancy was her first pregnancy. Her stillbirth experiences and her relationship with her living daughter were shared in this research. The time gap between the stillbirths to subsequent live births varied from 15 to 20 months.Table 1Contextual details of participantsParticipant number and pseudonymAge (years)Stillborn baby pseudonymStillborn baby's gestationLive baby pseudonymLive baby age (months)Partner/Husband pseudonymDiagnosis history1Ruth35Emma41William4StevenNo2Sharon32Oliver31Grace4KevinDepression3Sarah34Joseph34Jacob21DylanNo4Karen48Chloe32Shauna48JohnPTSD5Isabel28Ella (Mia twin sister)25Amelia30RichardDepression6Defne30Ufuk30Zeynep48MuratNo
Procedure {#Sec7}
---------
Information about the study was given to participants along with the consent form. All participants gave fully informed consent including explicit consent for their quotes to be used for research related publications. After completing a consent form and a demographics questionnaire, four open-ended questions were asked of participants (please tell me about your stillbirth experience; please tell me how you felt during your subsequent pregnancy; describe how you felt when you first saw your subsequent infant (first thoughts, first memories); Please describe your relationship with your child and has there been any change in your relationship with your child from birth to present?).
In addition, they were also requested to provide further explanations about their accounts to clarify some aspects of their shared experiences. Participants of the study were asked to write freely as much as, or as little as, they would like to write about their accounts. No time limit was given to participants. They were free to withdraw from the process at any time. The questions of the study were asked one at a time and further clarifications were requested in order to clarify the individual's experiences (e.g. if they used metaphors, they were asked if they can explain further what they mean by this). The questions of the interviews were asked over a period of time. In total 6--8 emails were exchanged between the researcher and the participant. For the purpose of the research, participants' personal and contextual characteristics were taken out and pseudonyms were used throughout, including participants, partners, babies and other children.
Participants were firstly asked about their stillbirth experiences; secondly, about their experiences of the subsequent pregnancy; thirdly their memories of giving birth to their living infant and finally their relationships with their living infants were examined. Each participant completed the email interviews within a time range varying from 8 weeks to 12 weeks. Once the email interviews were completed, debriefing information, including further information and relevant support services, were provided.
Results {#Sec8}
=======
An interpretative phenomenological analysis (IPA) was used in this study \[[@CR21]\]. Our analysis focused on how women make sense of their experiences of their stillborn baby and their relationship with their subsequent child, as well as the common ground shred by these women. The process set by Larkin and Thompson \[[@CR23]\] was taken as a guide and the following procedure was adhered to during the analysis of participants' accounts (see Table [2](#Tab2){ref-type="table"}). IPA analysis produced three principal themes: I). 'Broken Canopy'; II). 'How This Happened'; and III). 'Continuing Bonds. Overall 11 subthemes (Table [3](#Tab3){ref-type="table"}) were identified which are illustrated in the main features of the principle themes. The titles and labels of themes contain participants' own wordings where possible in order to stay close to the participants' own experiences. Some interpretation and discussion points were also highlighted within the themes' narrative. Then under the main theme headings all the themes were presented and discussed.Table 2Steps taken for the IPA1. Prepared transcripts for analysis (pseudonyms were given, identifying details were taken out and line numbers inserted). (By first author)2. Free coding followed up by a close, line-by-line analysis, was completed to understand each participant's concerns and claims. (By first author)3. Emerging themes were then established for each individual case in conjunction with regular supervisions. (By first & second author)4. Then, from the researcher's own understanding of theoretical frameworks in Psychology and from reflections from her own stillbirth experience, an interpretive dialogue was established and this was also highlighted in each transcript. (By first & second author)5. For each case a narrative overview, summarising emerged themes and the researcher's own interpretation and speculation for each case, along with the line by line coded transcripts, was established. (By first & second author)6. All participants' identified themes were presented side by side in a table for general visual overview. This was then used towards establishing the structure of the main and sub themes. (By first & second author)7. Then all participants' experiences were tabulated, this time according to the established structure and presented in a table in which the participants' contribution was indicated. (By first author)8. A narrative of women's experience, evidenced by extracts from participant's accounts, was then developed in conjunction with the established structure. (By first author)9. The final analysis and interpretations were also overseen by the 3th and 4th authors, and the overall findings in relation to perinatal loss, attachment and mental health literature were assessed.Table 3Women's experience of stillbirth and parenting experiences of subsequent infantI) Broken CanopyThe number of participants shared the experience 1. It cannot be true -- Baby with no heart beat(Pregnancy with a dead baby; Confronted by a dead baby; Choice and information)6 2. Questioned Self & othersa. The off script experiences of others6b. Others failure to acknowledge the loss6c. Changed view of self -- self is alone5 3. It cannot be true -- Baby with a heart beat(Consolation prize/Runners' up prize)6 4. Surreal Experiences(Joy and grief; Creating life like other women)6 5. Anxious parenting(Unrealistic expectations from self; Creating memories)5 6. Integrating death in lifeSelf-growth4II) How did this happen? 1. WhyIs the self the culprit?6 2. EmotionsAnger and despair5III) A Continuing Bond 1. My baby existed after allThey are brothers/sisters; We are a family; He/she is still my child6 2. Betrayal4 3. Longing and need to be in touch6
Theme I - Broken canopy 'Questioned self and the changed view of world --world may not be safe' {#Sec9}
-----------------------------------------------------------------------------------------------
One of the participants offered an image of a 'protective canopy', a symbol of the assumptions which we may make about the safety of the world. **It reflects a concern, shared by all our participants, about the puncturing of this canopy with the unexpected death of a baby*.*** Respondents described something akin to an **existential crisis**: the realisation - through the deaths of their infants - the 'safe world' was actually fragile and were vulnerable. The realisation that anything could happen to the canopy appeared to change their world view. The vulnerability of self and life became the new focus, as it is represented in **Karen'**s extract below:"*I think now that Chloe\'s death has left me with an almost constant awareness of the fragility of life, how quickly everything can change. Before Chloe died, a headache was just a reason to go to bed earlier, now I worry could it be something more serious. Now when friends are expecting babies, I feel great relief when their babies arrive safely. I don\'t have that blind expectation that all will be well. I don\'t trust doctors so much either****(KAREN)***"
This main theme, collectively shared by all the participants in various forms, presented itself in 6 sub themes: 'It cannot be true - baby with no heart beat'; 'Self and Others'; 'It cannot be true -- baby with a heartbeat'; 'Surreal Experiences'; 'Anxious parenting' and 'Integration of death in life'. Table [4](#Tab4){ref-type="table"} provides an overview for this large theme.Table 4Sub subordinate themes of broken canopySub subordinate themeQuoteShared experiences*1.1 It cannot be true - Baby with no heart beat*I couldn\'t breathe. I couldn\'t speak. Not only did I have to start processing this horrible information, but I had to experience it while still being pregnant...couldn\'t run. I couldn\'t fall to the floor. I had to hold up this big pregnant belly (SHARON)All participants shared their disbelief when they learnt the baby's heart stopped and their baby was no longer alive, although, they still looked pregnant and they still gave birth to their babies.*1.2 Questioned Self and Others*The death of a baby is so \"off script\". It\'s just not supposed to happen. And it taps into people\'s individual fears and discomfort (SARAH)This 'off -- script' experience then translated into failure of acknowledgement of such loss by others and the person being isolated from others as a result.*1.3 It cannot be true - baby with a heart beat*I know it\'s an odd observation to make, but I was really astounded by the fact she was breathing (ISABEL)The realisation of the broken canopy and the heightened awareness in danger and death appeared to leave the women surprised at having given birth to a living, breathing baby after all. Women collectively reported that they questioned their ability to create life and were prepared to face further adverse experiences.*1.4 Surreal experiences*I enjoyed seeing Grace on the screen at our many doctors\' appointments. Those were the moments I focused on her... But mostly, my thoughts and focus were on letting my hopes and dreams for Oliver go, and learning how I could incorporate his absence into my life (SHARON)A co-existence appeared to be linked with the surreal experiences and left mothers in a dilemma. All participants reported simultaneously experiencing opposite feelings -- joy and grief, were reported by all the participants.*1.5* Anxious parenting - (Unrealistic expectations from self; creating memories).This is all after she was born -going through the labour with her was a different thing entirely! I have described the change over from being in labour to having her born as two different worlds -diving off of a cliff only to land in a foreign land (SHARON)Heightened awareness of the imminent danger along with surreal experiences appears to influence mothers' relationship with their infants and their parenting experiences.*1.6* Integrating death in lifeIn fact, there had always been that fear of driving. It was difficult to imagine myself in traffic jam. But today I drive to work every day (DEFNE)The awareness of fragility of life and death itself appeared to bring a new authentic way of living. Life and death are not separate entities. Four out of six participants, articulated being able to find new ways of in engaging with life. Their focus appeared to move to the 'present' 'here and now' as described by Defne.
### It cannot be true - Baby with no heart beat {#Sec10}
The arrival of the baby who died in the womb appeared to bewilder women as the natural process of the pregnancy, and consequently the birth, was completed but without any living baby at the end.
**Women also seemed not to be sure what to expect from labour, especially about meeting the dead baby.** They were puzzled as their deceased babies were no different to a living babies in terms of the way they arrived and the way they appeared."*She was wrapped in a towel, like any other new born baby and handed to me. She was absolutely minute. Her face was bruised and there was a tiny trickle of blood coming from her nose and mouth. Her eyes were still sealed shut and she had no hair. She was still, clearly, meant to be in my womb. I held her and cried over her for a bit, before handing her to my husband who did the same****(ISABEL)***"
**Isabel**, in response to meeting her 17 weeks gestated baby stated that "it was still clearly meant to be in my womb". This summarised the out-of-place experience and also suggested a realisation that the full term was not complete and baby was too small to live and therefore there was nothing that could be done without the completion of the process.
In addition, half of the participants reported that they still held hopes for meeting a living child at the end of this natural process, regardless of the facts that they were given. **It appeared that the actual realisation of the baby's death did not happen until they met their baby in the flesh.**"*No-one had told me Chloe would be warm. I think deep inside without telling anyone I felt she was still alive. Later as she lay in the cot beside me I dozed off and when I woke I thought she had moved I screamed. The junior midwife came running in. I didn\'t tell her why I screamed****(KAREN)***"
**Upon meeting their stillborn babies, mothers appeared to instinctively want to take care of them.** Not being able to do so seemed to make mothers anxious, and particularly when they were separated from their babies, they seemed to realise the fact that they were not going to see their babies ever again."*We were able to hold him and spend some time with him, then they took him ...****When the hospital took Oliver away****, I felt empty. I wanted to know where he was going, who was going to take care of him, were they going to be careful even though he wasn\'t living. He was my child and I felt sick that I would never see him again. For me that was the beginning of my unyielding grief that he was no longer a physical part of me and I couldn\'t feel him anymore****(SARAH)***"
*'When the hospital took him away'* also suggests the mother's struggle in accepting her baby's death. This may suggest belief that he is still alive and also her helplessness. Furthermore, it emerged that women **were not informed about the options available** after their stillbirth experience and because of that **some women missed or almost missed their chances of being able to say goodbye to their baby**. For example, some only found out by chance about the possibility of keeping their baby overnight. While women were recovering from the demanding labour itself, these women were at the very same time also going through the realisation of their baby's death. **Karen** cited her gratefulness for being able to say goodbye to her baby, even if only by chance."*We found out on Monday, 3 days later that we didn\'t have to leave her there so we went and brought her home. It was so lovely to hold her and hug and kiss her and have her home. She was in her coffin in our bedroom until we buried her on Wed. Having Chloe home meant the world to us. Her big sister was able to hold her too and family visited and we felt we had 48hrs to tell her we love her. We never think about the fact we were not told we could take her home. That would be so upsetting. We are just thankful we found out we could bring her home****(KAREN)***"
**The importance of keeping a memory of the deceased baby becomes clear to most mothers at a later stage in their journey, as suggested by Ruth.**"*We were fortunate to have a local photographer from a local charity arrived at the hospital and took pictures for us. It seemed awkward at first but we are both very thankful to have these photos as they are the only ones we will have of our darling angel daughter****(RUTH)***"
Five out of six participants seemed to acknowledge the importance of being able to take the available opportunities (taking photos, taking the baby home etc.) later on. One of the participants (**Defne)**, however, was strongly encouraged by significant others 'to move on'- get on with life and hide her feelings and her longing to see her son. This may be one reason why she did not name her baby, had no physical memories of her baby, and relived her experiences on her own secretly.
### Questioned self and others {#Sec11}
**It emerged that the mothers' stillbirth experiences triggered fears and threatened their own assumed safe world.** Acknowledging this experience perhaps risked acknowledging the possibility of the fragility of someone's own canopy. Perhaps this was the underlying reason for others' unavailability for support and validation, and for their suggestions of dismissive strategies to the bereaved mothers.
Collectively all participants shared their **need to be recognised and acknowledged by significant others following their experience of loss.** However, validation of their feelings from others did not appear to be available or else was limited."*Only one of my friends said cry Defne. No matter what I will say will lessen your pain, but express your feelings to me- offered me a shoulder to cry on. I cried a lot that day, only to her... How well she understood my only need to be able to cry****(DEFNE)***""*Every now and then someone either a family member, friend or someone handling the burial arrangements would make a hurtful comment such as \"Don\'t worry, you\'ll have another baby.\" or \"Are you sure you felt fine? You didn\'t feel like anything was wrong?\" or \"It was God\'s will.\" None of this was helpful, because a) I don\'t want another baby. I want this one; b) if I didn\'t feel fine, I would certainly have rushed to the doctor or hospital!; and c) what little faith we had we were now questioning****(SARAH)***"
Women appeared to be **being isolated and alone in their experiences in this unknown new world.**"*I think it was tough, dealing with depression to be honest and my friends, the one or two I have in \[Place 1\], rarely showed up or text me, so they were little help sadly. My husband worked 60 hour weeks at the time, so when I did see him, he was exhausted himself****(ISABEL)***"
**The majority of mothers with their unmet needs of support appeared to be encouraged to keep their sorrow within as they felt lonely and isolated in their experiences.** There appears to be a changed view of self and others and feelings of isolation and loneliness."*It is wonderful to go to school with her everyday but I still cry when I think of my son but nobody knows it****(DEFNE)***"
### It cannot be true -- baby with a heart beat {#Sec12}
Women appeared to question **their ability to create, with the arrival of the dead infant.** However, until the arrival of the subsequent living infant, the concern that they may not be able to create or bring life was a strong possibility within the shattered unsafe canopy. Therefore, it appeared to be hard for the women to believe that they could have a living baby after all."*This may sound morbid, but I felt disbelief that I had actually given birth to a healthy child!****(SARAH)***"
**Women, seemed to try to separate the two different, but co existing, infants' places throughout their journey.** This breathing baby was not a substitute or consolation prize. Mothers refused to think that they were substituting their children with each other, as represented in Karen's account."*Other people seem to see Shauna as some kind of 'consolation prize' for Chloe\'s death. I find this so untrue and offensive. Giving birth to Shauna safely and rearing her did not heal my grief over Chloe. What it did do was give me a pressing reason to get up every morning. One child does not replace another. Each of my daughters has their own special place in my heart****(KAREN)***"
### Surreal experiences {#Sec13}
**Co-existence on the other hand appeared to allow mothers to define each child.** Sometimes the dead baby defined the existence of the living baby and sometimes the living baby defined the existence of the dead baby throughout the mothers' journey, as described by Sarah's extracts as shared in the co-existence:"*Often I felt I had to act like I was always happy and grateful in front of everyone else for their own relief and happiness about expecting Jacob. Mind you, I was thrilled to be expecting him, but that coincided with the fact that I was still grieving. The guilt didn\'t last long, because we came to see Jacob as a sign from Joseph that we should love another child as well. Perhaps that sounds a bit esoteric, but we believe that****(SARAH)***"
### Anxious parenting - (Unrealistic expectations from self; creating memories) {#Sec14}
When **Sharon** was asked to describe her relationship with her living subsequent infant, she described the loss and despair as a 'cliff', and that giving birth to a living child was like 'diving off from this cliff to land of an unknown -- parenting'. This analogy sums up the other women's experiences of how their parenting was influenced and shaped by their previous loss. Life and loss coexisted once again.
**Joy coupled with grief, and the shattered 'safe' world seemed together** to catalyse the women's constant worry of their living offspring's welfare. Mothers appear not to focus 'here and now', but rather their focus is either in the past or in their future worries."*There was so much worry. I had dreamed before she was born, of us being so relaxed and enjoying her baby time. I think now that my expectations were too high. But the worry for having Shauna did not turn out how I expected. I was surprised when one doctor commented that he couldn\'t understand why we worried so much about her. After Chloe died so suddenly, I felt it made perfect sense that we would be worried that something bad would happen to Shauna****(KAREN)***"
Furthermore, half of the participants appeared to be engaged in protective mothering activities, sometimes involving unrealistic expectations of self in order to protect the infant in an **unsafe** world."*My feeling toward my 'importance' to Daniel relate to my ability to provide for him in a way that no one else can. I really wanted to be able to breast feed him for at least 6 months. When this didn\'t happen, I felt less of a woman and almost helpless****(RUTH)***"
The constant awareness of the fragility of life and the emphasis on 'past' and 'future' motivated most of the mothers to anxiously collect mementos of their subsequent children. And this time, mementos are plentiful, unlike the first time. Now that they know how precious keepsakes can be, collecting them may be a way for mothers to protect themselves -- if this subsequent baby should die too, they will have more mementoes to keep and treasure, which would bring comfort as they grieve. This can make the risk of another loss more bearable to contemplate."*With Grace, if there was something that felt right for her, I bought it with the understanding that it is hers whether she ever used it or not. In that way, I was creating physical memories of her if we lost her****(SHARON)***"
### Integrating death in life {#Sec15}
**The awareness of fragility of life and death itself appeared to bring a new authentic way of living.** Life and death are not separate entities. Four out of six participants, articulated being able to find new ways of in engaging with life. Their focus appeared to move to the 'present' 'here and now' as described by Sharon."*Something beautiful that I experienced being pregnant after having a baby born still is that I treasured each moment that she was alive in me. Most people go through pregnancy anticipating the next steps - birth and life. Never having gotten to those steps with my son, I was able to build a relationship with my daughter in a unique way in utero. I was getting to know her and think about her in the moment rather than dreaming about the future****(SHARON)***"
Mothers appeared to be anxious about being able to protect their children in their shattered new world. They then tried to restore the broken canopy at any cost including sacrificing themselves via unrealistic expectations of the self. Heightened anxiety appeared to be a new focus in their relationship with their infant. However some women moved beyond their awareness of death and danger, and integrated the death and danger into their life, existence. Life and death together defined their existence**.** Gathering memories also requires that a mother soak up the present moment, which is another way to express an enhanced appreciation for this new baby's life.
In summary, it appeared that the realisation of the 'vulnerability of self' spurred participants into an appraisal process where their own abilities and others' availability to them were assessed. The view of an assumed - safe world changed and women appeared to feel isolated in their experiences. When women then turned to others for support they were urged to move on -- get on with life. This meant to mothers that their experiences were dismissed and rejected and this was deeply upsetting for women. Still, even as each mother turned toward her subsequent baby, she did not abandon her devotion to her deceased baby. The two babies appeared to co-exist with each other, as separate individuals, and the mother's experience of each coloured her experience of the other. For example, the arrival of a new baby could not be embraced by mothers as a completely new experience, as it was coloured by grief for the baby born still. The mother's experience of losing and grieving for her deceased baby rendered bittersweet her experience of welcoming and parenting the new, healthy baby. In kind, as the mother's emotional investment turns to the desire to create and nurture a new life. At the same time she worries she is betraying her deceased infant. This contradictory duality is articulated in powerful, simultaneously experienced opposing emotions like joy and grief, fruition and guilt.
Theme II -- How did this happen {#Sec16}
-------------------------------
Women collectively asked the question 'Why' in their accounts. One participant, Ruth described losing her baby as tragedy. This suggested extreme sorrow, as a consequence of a tragic flaw, however there was a meaningful ending."*When I am reminded of my daughter\'s tragedy I think to myself how lucky I am to have known her at all. I used the knowledge of her situation and took that forward with me during my pregnancy with William... Life is so precious and too many people take the ability to create life for granted (****RUTH)***"
Other women articulated answers for their quest in searching for answers as to why this had happened, and their account did not have self-blame as shared by Sarah and discussed under the co-existence subtheme.
### Why - Am I the culprit? {#Sec17}
It appears that when there were no meaningful answers to the question of why they were chosen to live without their children, most of the women (Five out of six participants) expressed anger in the form of self-blame towards self."*So, basically, we were left with absolutely no answers. I think that has been the hardest part in our process. Doctors tell us they don\'t like to have answers because it\'s less likely to recur. We like that. But it doesn\'t help in our understanding of what happened to our little boy. And it certainly adds to my anxiety that maybe it was something I did (****SHARON)***"
### Emotions - anger and despair {#Sec18}
Anger was also articulated by almost all participants, except Ruth, toward various significant others including family, friends, hospital staff, God and the baby."*Everybody said if this happened later it would have been worse, what happened was better than what would have happened if this child born with disabilities. This made sense but it did not make me feel better. Even it made me angry****(DEFNE)***"
Isabel, in response to a consultant's dismissive statement for her constant worry during her subsequent pregnancy, stated that"*I could have chucked a chair at his head. I pointed out to him, rather curtly, that I had buried two children and that I didn\'t plan on doing it again and if he\'d been through what my husband and I had been through, he wouldn\'t be asking such a dumb-ass question****(ISABEL)***"
Anger was also expressed towards God, and the baby, in participants' Sharon's and Karen's accounts."*I could not touch him. When asked prior to his birth, I had told everyone I was going to hold him. But when he came out I felt differently. All I kept thinking was, \"That\'s not him\" I knew the real essence, the true being who had been my little boy, was not in that body. My baby was gone. I said I had held him for eight months. I wasn\'t going to hold him when he wasn\'t there****(SHARON)***""*I tried going to Church but gave that up quite soon as I was so angry with God****(KAREN)***"
Women collectively also talked about their despair and helplessness in the situation that they were in while articulating their realisation that there is nothing they can do or undo to change the circumstances; this is represented by Ruth's extract."*I spent the first few days just completely numb, like a robot, coordinating and planning her funeral services. I felt the need to make sure she received the best she could, since there wasn\'t anything else I could do for her****(RUTH)***"
In summary, sooner or later, the arrival of a sudden 'end before a beginning' appears to come as a shock to all women. Then they appeared to continue to question the self and others, while looking for the reasons why they had to go through such experiences. All participants expressed despair, while anger towards self and others was shared by most of the participants (Five out of six participants).
Theme III - A continuing bond {#Sec19}
-----------------------------
A final theme emerged around the enduring relationship with the deceased child, which was articulated by all the participants. As discussed in the summary of the "Broken Canopy" theme, the mother's enduring bond was also observed in the co-existence of her deceased and living infant. This was also observed in the co-existing relationship between deceased and living infants. Three subthemes emerged under this theme as follows:
### My baby existed after all {#Sec20}
The death of an infant before birth giving rise to hardly any memories seems to complicate the natural bereavement process. Women are faced with a baby's death before fully realising this baby's existence. It appears that this realisation becomes clearer only with the arrival of the new baby. This collectively shared experience was articulated in Sharon's excerpt."*There is the real child, and the one we have created in our minds. Understanding this was a beautiful and helpful thing for my relationship with Oliver. It made part of him still exist for me. Accepting this allowed me to continue to know him as my child. I just had to come to terms with the fact that I would never get to see who he was as compared with my creations****(SHARON)***"
### Betrayal {#Sec21}
As mentioned under the "Broken Canopy" theme, the majority of participants (4 out of 6), felt that they have betrayed the deceased infant when they became pregnant again and subsequently gave birth to a living infant. Women's natural desire to have children appeared to contradict the desire to have children appeared to contradict the desire to feel connected to the baby who died, as illustrated by Sarah."*It was very complicated. I felt terribly guilty, as if we were already forgetting Joseph. I\'m sure others judged the fact that we conceived right away, but our doctor recommended it and I was already 33. Even when I discovered the positive pregnancy test, I remember calling Dylan and just feeling scared and nervous. It was difficult to enjoy the pregnancy and all the joys of expecting\--first kicks, ultrasound photos, etc. We were constantly fearing for the baby\'s life****(SARAH)***"
It also appears that the majority of women at a later stage in their journey reflected on how they dealt with the process and regretted the missed opportunities afterwards, as articulated by Sharon."*She presented me with a card with the footprints in and photos of her. I was very grateful for her doing that, although I think back now and wish I\'d have done it myself. I was her Mum after all****(SHARON)***"
### Longing and need to acknowledge a continuing bond {#Sec22}
Women expressed a longing for their infants and the need to stay connected somehow."*'My husband and I felt incredibly lonely in the sense that we had these empty, aching arms that should be more than filled with two babies****(ISABEL)***"
Women chose different ways to be in touch with their baby and their memories:"*Ok, keep the questions coming. I am glad to be purging all of this. Sometimes I go weeks without talking about Joseph****(SARAH)***"
And some stayed in touch with their infant via involving themselves in activities in their child's memory, supporting families going through similar experiences or taking part in research in the area of stillbirth experiences."*I just love to talk about my daughter; it helps to 'keep her alive' in my heart****(RUTH)***"
One of the participants said that although their infant is not living anymore, the child is still part of their family, including the subsequent children. This loss is also the whole family's loss including the subsequent children."*I walk around thinking I should have two sons on either side of me. My husband says we are blessed with our son Jacob because we lost our first son Joseph, but I still feel that they are brothers who should be together right now, playing, getting into trouble, getting ready to start nursery school, etc. I think of myself as a mom to two boys, but no one else sees me that way****(SARAH)***"
As discussed in "Broken Canopy" it appears that the co-existence of the dead and living babies enables the mother to be connected with her deceased infant for whom she still feels a longing and devotion. **Isabel** reported explicitly how she connected with her infants via sensory experiences with her living infant."*Coped with the guilt of devoting all my time and attention to Amelia by doing certain things. Probably sounds weird, but there are times when I can \'smell\' them. All babies have a particular scent and so did Ella before she died. From time to time, I can smell her and I always say hello to both her and Mia****(ISABEL)***"
In summary, mothers appeared to be initially occupied with making sense of their baby's death. The recognition of the baby's existence is something articulated in all mothers' accounts at a later stage. At the same time mothers expressed longing for the baby who died and the need to acknowledge their maternal bond. Betrayal of the deceased infant was also expressed when the mother experienced joy. They reflected on their missed opportunities, such as spending more time with and holding this baby, and keeping memories alive so they can continue to feel connected with this child.
Discussion {#Sec23}
==========
Women's accounts revealed that the experience of stillbirth is a process where women **re-visit** the experience and reflect their experiences **throughout other life events such as the arrival of a new baby.** The experience of stillbirth appears to **influence the relationship with the subsequent infant and parenting.** Further discussion and interpretation points, for the three main principle themes follow:
Broken canopy {#Sec24}
-------------
Women collectively appeared to question their sense of mastery in the world, and the foundations upon which they build their lives following their stillbirth experiences. It is plausible that an existential crisis as discussed by Yalom \[[@CR24]\] appeared to be evident in participants' accounts, particularly, in debating thoughts of existence vs nonexistence and the fragility of life. Similarly Janoff -- Bulman \[[@CR25]\] discusses how the death of a loved one shatters individual's world view and core assumptions about self. Women's increased awareness of death and their questioned ability to be able to cope appeared to be translated into **constant awareness and anxiety in** the **subsequent parenting experience**. This finding is in line with the emerging literature for parenting following perinatal loss \[[@CR18], [@CR26]\]. Yet some women reported an integration of the baby's death into **their lives, a self-growth** following their trauma experience, as discussed by Yalom \[[@CR24]\], Linley and Joseph \[[@CR27]\], and Davis and Nolen-Hoeksema \[[@CR28]\]. Furthermore, this authentic, enriched perception appeared to enhance the **mothers' relationships with their subsequent children. This finding is similar to** the findings of a recent qualitative self - growth study following stillbirth \[[@CR29]\] and the findings of Cacciatore \[[@CR30]\] (2010) and Lichtenthal and her colleagues \[[@CR31]\] (2010) following the death of an infant. The current study extends this understanding to the subsequent parenting experiences of mothers and supports the emerging findings in relation to the influence of self-growth on parenting \[[@CR32]\] (O'Leary & Warland, 2012).
From an existential point of view \[[@CR24]\] women were faced with the possibility of their inability to be immortal by generating life and so had the urgency to try for other babies. At the same time women also needed to grieve and continue to feel connected with their stillborn babies and their memories. This desire to have more children in order to generate life, seemed to have left women with the dilemma of worrying about betraying their deceased infant \[[@CR30], [@CR33]\].
The fact that mothers questioned both themselves and others in response to such an existential threat can also be examined from the perspective of attachment theory \[[@CR34]\]. Bowlby suggests that attachment behaviour is activated under a threat and individuals then engage in support-seeking behaviour. People with insecure attachment, who had inconsistent or lack of security-providing-interactions with attachment figures, are expected to have doubts about the effectiveness of available support, and will use other secondary strategies (such as, the use of deactivating strategies to idealise and normalise relationships, no or lack of memory in relation to early care experiences \[[@CR35]\], or self blame \[[@CR36]\]). The lack of validation of feelings and rejection by others may is an issue for individuals with insecure attachment style and could lead to self-worthlessness and self-blame. All participants collectively wanted their experiences to be acknowledged. Stillbirth, such an 'off-script' experience, appears to threaten other people's own assumption of a safe world, making them less likely to validate the mothers' experiences and feelings. This in itself appears to isolate women in their bereavement process and forces them to hide or deny their feelings. Mothers found it hard to deal with others' dismissive approach (e.g. suggesting 'moving on' or reminding women to be grateful for their living baby). This rejection also appeared to cause one woman to criticise herself. For example, this participant stopped seeking medical and social help while going through grief, depression, and the demanding needs of caring for her subsequent newborn baby. Being critical of oneself and a lack of social support are both identified risk factors for prolonged grief or delayed grief reactions. Acceptance, however from the wider community helped women to embrace their experience. These findings concur with current research on the importance of social support \[[@CR37]--[@CR40]\]. In addition, the lack of validation of feelings and rejection by others appears to confirm the prior rejection of caregivers and results in the mother's experience of disenfranchised grief as discussed by Nichols \[[@CR38]\].
What was also striking in the mothers' accounts was the co-existence of contradictory powerful feelings, described by one of the mothers as 'surreal experiences'. Women reported experiencing joy and at the very same time disbelief, when they gave birth to a living, breathing baby. The joy was also coupled with their grief and longing for the dead infant. The co-existence between the dead baby and subsequent living baby was also reported during the subsequent pregnancy from conception to birth and was even present in mothers' parenting experiences of the subsequent baby. Mothers appeared to define their subsequent baby's existence according to their bereavement experiences and they seemed to be able to process the existence of the stillborn baby, who they had never seen alive, via their interaction with their living infant. This co-existence also enabled women to feel a continuing bond with the longed-for infant. However, the women were also aware of each infant's individual place and existence. Although they acknowledged that one of their infants was not alive, they fought for their infants' separate places, especially when the outside world appeared to dismiss or ignore this independent existence ('they are like brothers'; 'second child is not a consolation prize').
In addition, it can be speculated that the co - existence of the two infants also helped women in their grieving process, however, this hypothesis needs further investigation in future studies that include controls of women who do not have any living children. The study's findings regarding co-existence does extend the current "replacement child" and "vulnerable child" debate as discussed by Turton and colleagues \[[@CR17]\], while bringing attention to the connection between the deceased and living infant and in some ways, the co-existence being therapeutic for the mother.
How this happened {#Sec25}
-----------------
Whilst mothers were facing the new existence of a subsequent child, they all questioned their ability to cope and others' availability for them during their journey. This inevitably gave rise to the question 'why', and in response, women expressed anger towards self and others, including significant others. Anger towards self, and others were reported by almost all participants. Questioning of the self and self-blame as part of the anger process was also observed, similar to the findings of another qualitative study \[[@CR31]\]. Only one of the participants did not express anger or self-blame towards self and others. This perhaps was due to the way in which she conceptualised the loss, and integrated the death into her life, or the way in which her experience was also acknowledged by her close community, unlike other participants' experiences. Perhaps the acceptance and validation of their experiences and their feelings at an earlier stage contributed to the person's own acceptance of the situation without turning the woman's anger towards herself or others. Further research focusing around the need of validation and acceptance of mothers' experiences may expand this speculative point. However, it is evident both in this study, and other similar studies \[[@CR41]--[@CR43]\] that support from significant others was sought after stillbirth and needed for the experiences and feelings to be validated.
A continuing bond {#Sec26}
-----------------
Another overarching theme was about continuing the bond with the dead infant. As discussed by Klass and colleagues \[[@CR44]\] the need to be in touch with, and longing for, their deceased infant was shared by all the participants. The adaptive value of this continuing bond has been discussed in the literature \[[@CR45]\] and this was evident in the accounts of mothers of the current study. The connection with the baby was sometimes achieved via a living baby, by holding on to their few memories, experiencing co-existence, engaging in activities like research, or supporting other families. This theme also emerged in another study where women's accounts of stillbirth experiences were examined in relation to self-growth \[[@CR29]\].
According to attachment theory, the need for a continuing bond can be an indication of failure to integrate the death of a loved one, and individuals can be classified as having an unresolved state of mind with loss. Some mothers in their accounts refer to their deceased infant as 'gone', which could imply "just left, not dead" or "had been left at the hospital". Using this euphemism could suggest a disorganised belief in relation to loss and unorganised state of mind. However, the unresolved category should only be given when there is a disorganisation or disorientation in discourse or reasoning by the individual during the discussion of traumatic events (e.g. loss, abuse) \[[@CR46]\]. Although "unresolved griefs" was not explored in this study it is important to understand this dynamic in order understand and appreciate the adaptive value of continuing bonds.
In the current literature, there are contradictory findings in terms of the adaptive effects of continuing psychological and emotional bonds with the deceased loved ones. Klass and Walter \[[@CR47]\], Field \[[@CR48]\] and more recently, Field and Filanosky \[[@CR49]\] identify continuing bonds as either internal or external continuing bonds (CB). Their analysis, inspired by attachment theory, revealed that external CBs (illusions and hallucinations) were positively correlated with guilt and feelings of responsibility for the death, whereas internalized CBs (use of deceased as an autonomy - fostering secure base) were negatively associated with identified risk factors as well as uniquely associated to personal growth. In the current study it appears that the grief process becomes complicated in the stillbirth experience, as death occurs before life outside the womb and there are so few memories to hold dear, which engenders a feeling as if the baby never existed. Therefore, it stands to reason that supporting mothers in being able to acknowledge the baby's life and death will help them accept their infant's existence while knowing they are not living. Perhaps this dilemma is one of the reasons for worsening mental health problems in these mothers, such as continuing depression and PTSD \[[@CR17], [@CR50]\]. It can be speculated that the issue in the current debate about the link between PTSD and seeing and holding the deceased baby lies in the existence of externalised continuing bonds between mothers and their deceased infant as discussed by Field and Flonosky \[[@CR49]\]. Further research is required to investigate this possible relationship.
Importance of findings {#Sec27}
----------------------
The findings of this study provide **an insight into the stillbirth experience** of mothers and its meaning to them with **an existential focus.** It highlights the dilemmas and difficult decisions that women face in their experiences. It also provides **evidence for the importance of a continuing maternal bond, how the stillbirth experience influences the mother's subsequent pregnancy and parenting.**
This study reveals that the **mothers' struggles to accept the existence of her baby while being aware of the non-existence of her baby, as she has no shared or past memories other than those of the pregnancy and birth.** It can be speculated that the lack of memories and opportunities to nurture her baby perhaps then complicates the grief process \[[@CR51], [@CR52]\] and may keep the mother in denial or extended the recovery period where symptoms of depression and PTSD.
Every mother in this study saw her baby, although not everyone chose to hold her baby. None of the mothers regretted seeing her baby. Only one research group suggests a link between PTSD and seeing and holding a deceased baby \[[@CR15]\]. However, in the current study, it seems that **meeting with the dead baby** actually **was a crucial point at which women started processing their grief.** Only from this point onward was there a full acknowledgment of their baby's death, unlike the experience of pregnancy with the dead baby and giving birth to the dead baby. These findings are in parallel with the recent related findings and advice \[[@CR5]--[@CR8]\] that mothers benefit from health care practitioners assuming they feel a maternal bond and then supporting them around spending time with their babies.
Although **seeing the deceased baby seems to facilitate the grief process, the established strategies of each individual were important while they were dealing with the emotional aftermath of meeting their dead baby.** For example, a mother with dismissive strategies or mothers with avoidant attachment styles, may find it difficult to process such direct contact. Although this is a speculative point it is, however, important to note the importance of **individual differences in dealing with stressful situations** when providing efficient guidance in the management of stillbirth. Therefore more research should be carried out to understand individual differences in dealing with stillbirth experience and this should then inform the relevant guidance It is also important to note that participants did not receive clear information about their options for spending time with their stillborn babies. This could be because of a hesitant attitude of the staff due to the current, mixed guidance. **Therefore a clear and unified guidance is essential for better stillbirth bereavement care** \[[@CR53]\].
Furthermore, mothers' awareness of danger and heightened anxiety, along with their unmet support needs, were present during their subsequent pregnancy and their parenting of a subsequent infant. These findings are in line with the available literature \[[@CR12], [@CR54]--[@CR56]\]. Anxious parents can become controlling and critical, may experience difficulty in bonding, and subsequent infant attachment can be disorganised \[[@CR15]\]. This has further implications for the subsequent infants' adult life, including the possibility of anxiety and depression disorders \[[@CR57]\].
Limitations {#Sec28}
-----------
Due to the nature of the qualitative design, the sample of the study is small. Therefore, generalisations of the findings are limited. In addition the women who participated in the study were recruited via web based social networks thus it is only representative of mothers' who have actively sought more public venues to reflect and share their stories and remain in touch with their infant.
Conclusions {#Sec29}
===========
First of all, the findings of this study inform the professional practice for pre and post-care of mothers who experienced stillbirth. It provides a better understanding of mothers because it explains the meaning to the mother of a stillbirth. Particularly relevant for psychological support services is that emphasis should be placed on the acceptance of the dead baby and co-existing experiences (e.g. joy & grief; betrayal & fruition). Including the deceased child in the family and its narrative may also allow the mother to integrate her stillborn baby her into her life and allow her to realise that her baby existed but is no longer living. This may allow women to grieve, honour her maternal bond, and stay in touch with their baby's memory. Mothers' need for a continuing bond should also be recognised and the unmet validation needs of women should be part of the psychological support process. Issues around anxious parenting should be expressed and addressed appropriately, taking into account individual needs. The findings from this study could also inform public health authorities regarding the need for awareness of stillbirth and a better stillbirth bereavement care (e.g. available information, support in difficult decisions) and that individual differences in response should be taken into consideration. Particular attention should be given to the isolation that women experience due to their 'off-script' experiences and disenfranchised grief. The findings of this study also had personal implications in that the researcher had a chance to reflect on her own experiences and remain in touch with her own experiences.
A. Meltem Üstündağ -- Budak, Michael Larkin and Gillian Harris contributed equally to this work.
**Competing interest**
The authors declare that they have no competing interests.
**Authors\' contributions**
AMuB conducted the interviews and data analysis and prepared the manuscript. ML supervised the data analysis and drafting the manuscript. GH and JB overviewed the study and the drafting process. All authors read and approved the final manuscript.
The authors thank to Birth Trauma Association (BTA), UK and Share US (Pregnancy and Infant Loss Group) and all participating mothers for taking part in the study.
| {
"pile_set_name": "PubMed Central"
} |
In the article cited above, the addition of \"^b^\" annotation would be needed in Table 5 of the main text.
^b^In case of CURB-65, the variables are as follows: age \< 65 years, absence of chronic cardiovascular disorder, absence of chronic lung disease, absence of CNS disorders, and absence of chronic renal disorders.
Hence, meaning of the text would need the correction. The correction text from eighth line to thirteenth line of the first paragraph, page 769, is as follows: \"The propensity to score low on the CURB-65 was decreased in the patients aged 50 to 64 (*p* \< 0.001), in those without chronic renal disease (*p* \< 0.001), and in those without a CNS disorder (*p* = 0.024). Moreover, a high CURB-65 score was more likely in patients without chronic lung disease (*p* = 0.014).\"
We apologize for any inconvenience that this may have caused.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#S1}
============
The adaptive immune system must distinguish between self and non-self in order to provide protection from pathogenic challenges while sparing the organism's own tissues. Recognition of B7 ligands (CD80 and CD86, also known as B7-1 and B7-2, respectively) by co-stimulatory CD28 and co-inhibitory CTLA-4 (cytotoxic T-lymphocyte-associated protein 4, also known as CD152) receptors plays a critical role in regulation of effective self versus non-self discrimination. CD28 signaling is required for optimum proliferation and function of effector T cells, whereas CTLA-4 plays a critical role in negative regulation of immune responses, as it is required for turning off effector T cell signaling and regulatory T cell (Treg) development and suppressive functions. These opposing immunomodulatory roles of CTLA-4 and CD28 are of considerable clinical significance. CTLA-4 was the first immune checkpoint receptor targeted for cancer immunotherapy, and the anti-CTLA-4 antibody ipilimumab is used in the clinic for treatment of advanced melanoma ([@B1]). CD28 co-stimulatory function is also relevant for cancer immunotherapy, as chimeric antigen receptors (CARs) containing CD28 cytoplasmic regions have been shown to induce efficient T cell effector functions ([@B2]). However, targeting CD28 with the superagonistic monoclonal antibody TGN1412 was a tragic failure, when administration of the antibody during a phase I clinical trial induced severe systemic inflammatory responses in healthy volunteers ([@B3]). Therefore, a comprehensive understanding of expression patterns, signaling pathways, and functional roles of CD28 and CTLA-4 on effector and Treg subsets can have significant medical impact.
CD28 and CTLA-4 recognize their B7 ligands in the context of the cell-to-cell interface, termed the immunological synapse (IS), formed between a T cell and an antigen-presenting cell (APC). Receptor ligation at the IS leads to accumulation of interacting molecules at different regions of the synapse, forming distinct molecular patterns known as supramolecular activation clusters (SMAC) ([@B4]--[@B6]). The canonical mature T cell IS consists of a central SMAC (cSMAC) containing TCR (on the T cell) and pMHC (on the APC) molecules, surrounded by the peripheral SMAC (pSMAC) containing LFA-1 (on T cell) and ICAM-1 (on APC) adhesion molecules as well as F-actin. The outer ring of the IS, known as the distal SMAC (dSMAC) contains molecules with large ectodomains, such as CD45 and CD43. The SMAC regions contain smaller microdomains, known as microclusters ([@B7]). The IS is highly dynamic, with movement of TCR microclusters toward the center of the synapse, where they undergo endocytosis. Antigen recognition under physiological conditions does not always result in formation of this canonical IS structure; nevertheless this model provides a useful framework for understanding spatial dynamics of molecular interactions at the interface between T cell and APC membranes. The IS is the main site of immune receptor triggering and recruitment of signaling intermediates, leading to signal initiation and integration. B7 ligand recognition leads to distinct localization of CD28 and CTLA-4 receptors at the SMAC, modulation of cytoskeletal dynamics as well as recruitment of protein kinase C (PKC) isoforms to the IS. The effect of B7 ligand recognition on the IS dynamics is cell type specific, with effector T cells and Tregs displaying different CD28 and CTLA-4 localization, leading to differential recruitment of PKC-θ and PKC-η to the effector T cell and Treg synapses. This review presents a brief outline of the roles of CD28 and CTLA-4 in the immune system, followed by a more detailed discussion of CD28 and CTLA-4 localization patterns in the IS, and the consequences of B7 ligand recognition on IS structure and stability in T effector and Tregs.
B7 Ligand Recognition: Structural Features and Expression Patterns {#S2}
==================================================================
B7-1 and B7-2 (CD80 and CD86) molecules share a similar structure, consisting of one membrane-distal variable domain-like and one membrane-proximal constant domain-like immunoglobulin superfamily (IgSF) domain. Purified CD80 crystallizes in a dimeric form, and undergoes spontaneous homodimerization in solution ([@B8]), whereas CD86 crystalizes as a monomer ([@B9]). The two different oligomeric states of B7 were also observed using Forster resonance energy transfer (FRET) analysis on the surface of APCs, with CD80 present on the cell surface mainly in the form of dimers, and CD86 being monomeric ([@B10], [@B11]). CD80 and CD86 are expressed on dendritic cells (DCs), macrophages, and B cells, with CD86 displaying higher constitutive expression and more rapid upregulation after activation. B7 molecules are also expressed on activated mouse and human effector T cells ([@B12]--[@B14]). CD80 and CD86 bind to CTLA-4 with significantly higher affinity than to CD28. CD80 is a stronger ligand, with K~D~ 0.2 μM for CTLA-4 and 4 μM for CD28 interaction, whereas the K~D~ for CD86 binding to CTLA-4 is 2 and 20 μM for CD28 ([@B15]).
CD28 monomers consist of a V-like IgSF extracellular domain, transmembrane regions, and a short cytoplasmic tail with no enzymatic activity. CD28 is expressed on the cell surface as a glycosylated, disulfide-linked homodimer of 44 kDa chains. In adult humans, CD28 is constitutively expressed on approximately 80% of CD4+ and 50% of CD8+ T lymphocytes. Loss of CD28 expression, most marked in the CD8 compartment, has been observed in humans during aging and autoimmune diseases ([@B16]--[@B18]). CD28 is expressed on all mouse T cells, and it is not downmodulated during aging ([@B19]). Repeated *in vitro* antigenic stimulation ([@B20], [@B21]) and exposure to common-γ chain cytokines or type I interferons ([@B22]) leads to downregulation of CD28 expression on human T cells. However, *in vivo* antigenic stimulation has been reported to increase CD28 surface levels on mouse T cells ([@B23]).
CTLA-4 shares structural similarity with CD28, forming homodimers of V-like IgSF monomers. CTLA-4 contains a 36-amino-acid-long cytoplasmic tail with no enzymatic activity. CTLA-4 is not expressed on the surface of resting effector T cells ([@B24], [@B25]), but is expressed constitutively in Tregs ([@B26]) under control of Foxp3 and NFAT ([@B27]--[@B29]). In both conventional T cells and Tregs, surface CTLA-4 is continuously endocytosed via a clathrin- and dynamin-mediated pathway, and recycled to the plasma membrane ([@B30]--[@B34]). Activation of effector and Tregs leads to upregulated levels of CTLA-4 on the cell surface. CTLA-4 internalization is mediated by the heterotrimeric adapter protein AP-2 ([@B30], [@B34], [@B35]) \[regulation of CTLA-4 trafficking is the subject of an excellent recent review in Ref. ([@B36])\], whereas CTLA-4 trafficking from the trans-Golgi network to the cell surface involves formation of a multimeric complex consisting of transmembrane adapters TRIM and LAX, as well as small GTPase Rab8 ([@B37], [@B38]). CTLA-4 present in recycling endosomes is protected from lysosomal targeting through interaction between LRBA protein (lipopolysaccharide-responsive and beige-like anchor protein) and CTLA-4's tail region ([@B39]). Since its lysosomal degradation involves interaction with another clathrin adaptor complex AP-1 that binds to the same tyrosine-based motif (Y201) of CTLA-4 as LRBA ([@B35]) (the interaction motifs in CTLA-4 cytoplasmic region are summarized in Figure [1](#F1){ref-type="fig"}), it has been suggested that the binding of LRBA may prevent interaction with AP-1 and thereby protect the protein from degradation ([@B39]).
![**Molecular interactions in B7 ligand recognition**. **(A)** Schematic representation of CD28 and CTLA-4 binding to the B7 ligands. **(B)** Schematic representation of the cytoplasmic regions of CTLA-4 (top sequence) and CD28 (bottom sequence). Known interaction partners of CTLA-4 are shown above and of CD28 below the alignment, and the motifs implicated in these interactions are color coded as indicated. Figure based on Hou et al. ([@B40]), Isakov and Altman ([@B41], [@B42]), Margulies ([@B43]), Schneider and Rudd ([@B36]), Sharpe and Freeman ([@B44]), and Stamper et al. ([@B45]).](fimmu-07-00024-g001){#F1}
Both CTLA-4 and CD28 rely on the amino acid motif MYPPPY in the vicinity of Y139 in human CTLA-4 and Y123 in CD28 for binding to the B7 proteins ([@B46]--[@B48]). Importantly, despite the identical amino acid sequence of the interaction site, CTLA-4 and CD28 are capable of effectively discriminating between B7 proteins. A key study from the Allison lab ([@B48]) reported that the binding of a B7 ligand was critical for the concentration of CTLA-4 at the IS and contributed to the concentration of CD28, and that CD86 was a preferred ligand for CD28 and CD80 for CTLA-4. Antigen-pulsed B cells expressing CD80 effectively concentrated CTLA-4 at the synapse. Furthermore, in synapses formed by B cells expressing only CD80, there was evidence for competition between CTLA-4 and CD28 for ligand binding, as CD28 accumulation was reduced even further when CTLA-4 was present at the IS. Conversely, peptide-pulsed B cells expressing only CD86 strongly increased the accumulation of CD28 at the synapse, but failed to recruit CTLA-4 ([@B48]).
CD28 in Regulation of the Immune Response {#S3}
=========================================
CD28 is the prototypic co-stimulatory molecule, and CD28 ligation leads to enhanced cytokine production, cell survival, and proliferation of effector T cells. The critical role of CD28-mediated signaling in optimum T cell responses is demonstrated by the T cell effector functions afforded to second-generation CARs containing cytoplasmic regions of CD28 and CD3ζ, but not by first-generation CARs lacking CD28 sequences ([@B2]). The cytoplasmic region of CD28 contains two main signaling motifs (summarized in Figure [1](#F1){ref-type="fig"}): a proximal YMNM motif and a distal proline-rich PYAP motif ([@B49]). The YMNM motif mediates phosphatidylinositol 3-kinase (PI3K) binding ([@B50]--[@B52]), leading to Akt activation; YMNM can also bind to GRB2/GADS adaptor proteins ([@B51], [@B53]) and the PYAP motif binds to Lck ([@B54]), filamin A, and GRB2/GADS ([@B53], [@B55]). The YMNM motif is followed immediately by another poly-proline motif PRRP, reported to bind the kinase Ikt ([@B56]). Analysis of knock-in mutant mice revealed that the PYAP motif is critical for IL-2 production and proliferation *in vitro*, as well as for *in vivo* antibody production and germinal center formation ([@B57]), whereas YMNM plays a role in augmenting T cell proliferation ([@B58]). Interestingly, knock-in T cells with both their YMNM and PYAP motifs mutated display less severe activation defects than CD28-deficient T cells, suggesting some functional role for the PRRP motif and/or yet unidentified cytoplasmic sequences.
CD28 is required for the thymic generation and peripheral maintenance of a functional Treg population. CD4+ Foxp3+ Tregs are key negative regulators of T cell-mediated immunity and are required for the control of spontaneous responses to self through several mechanisms ([@B59], [@B60]). Contact-mediated suppression relies on CTLA-4 interactions with its ligands and is discussed in detail below. Bystander suppression is mediated by suppressive cytokines, mainly IL-10 ([@B61]) and TGF-β ([@B62]) produced by activated Tregs, and by induction of cytokine starvation in target cells by IL-2 clearance ([@B63]). B7 ligand recognition plays an important role in Treg development and function, summarized in Table [1](#T1){ref-type="table"}. In CD28-deficient NOD mice, the percentage of peripheral Tregs is strongly reduced ([@B64]). Similar reductions are observed in NOD mice lacking both CD80 and CD86, leading to the conclusion that the B7--CD28 interaction is required for the formation of the full Treg repertoire. The reduction in the percentage of Tregs in NOD mice treated with B7-blocking CTLA-4-Ig correlates with a higher incidence of spontaneous autoimmune diabetes ([@B64]). Subsequent analysis revealed that Treg deficiency in CD28^−/−^ mice can be traced back to thymic development. The percentage of Treg precursors among thymic CD4 single-positive cells is significantly reduced in CD28^−/−^ mice as well as in NOD mice injected with anti-CD80 and CD86 antibodies ([@B65]), and in B7 double knockout mice ([@B66]). Peripheral homeostatic expansion of Tregs -- but not effector T cells -- in normal syngeneic hosts is also strongly suppressed by anti-CD80 and CD86 antibodies ([@B66]). A mechanistic explanation for the thymic requirement for CD28 was proposed by Tai et al. ([@B67]) who examined the consequences of CD28 deletion in a TCR-transgenic model. Mice expressing the AND TCR and its agonist ligand, a pigeon cytochrome *c* peptide, were found to effectively induce thymic Tregs only in the presence of CD28. This means that while a strong selection signal through TCR is indeed required ([@B68]), it is not sufficient for the full initiation of the agonist selection program leading to the generation of Tregs, and that a co-stimulatory signal from mTEC-expressed B7 molecules through CD28 is also required. It is noteworthy that a small proportion of regulatory phenotype T cells were still generated in the absence of CD28 but these cells lacked suppressive capacity ([@B67]). Earlier data from the same group indicated that CD28 is also required for the deletion of thymocytes by negative selection ([@B69], [@B70]).
######
**B7 ligand recognition in Treg synapse formation and suppressive functions**.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Surface interactions Biological significance for Tregs Reference
-------------------------------- ------------------------------------------------- --------------------------------------------------------------------------------------------------------
CD28--B7 Motility\ Lu et al. ([@B160]), Thauland et al. ([@B163])\
Tonic signals necessary for survival Zhang et al. ([@B71])
CD28--B7 Antigen recognition\ Apostolou et al. ([@B176]), Jordan et al. ([@B68]), Knoechel et al. ([@B177]), Walker et al. ([@B178])
Motility stop signal
Lu et al. ([@B160]), Thauland et al. ([@B163])
TCR--pMHC Synapse formation and stabilization Onishi et al. ([@B161])
Activation Schmidt et al. ([@B179]), Zhang et al. ([@B71])
LFA-1--ICAM-1 Proliferation Walker et al. ([@B178]), Zheng et al. ([@B155])
Surface accumulation of CTLA-4 Catalfamo et al. ([@B170])
CTLA-4--B7 Synapse stabilization Onishi et al. ([@B161]), Zanin-Zhorov et al. ([@B151])
TCR--pMHC
LFA-1--ICAM-1 Contact suppression Kong et al. ([@B152]), Qureshi et al. ([@B146])
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
A study using mice in which CD28 was selectively deleted in cells expressing Foxp3 (Cd28-ΔTreg), reported only a minor decrease in the percentage of thymic Treg precursors ([@B71]). This is in line with previous observations that CD28 is involved in the generation of early, Foxp3-negative Treg precursors ([@B72]). However, in stark contrast to earlier studies, homeostatic expansion of Tregs in the periphery was reported to occur independently of CD28 ([@B73]). Tregs from CD28-ΔTreg mice displayed reduced suppressive capacity, and consequently CD28-ΔTreg animals developed spontaneous autoimmunity ([@B71]). Costimulation through CD28 is required for *in vivo* expansion of Tregs in the presence of TCR stimulation and IL-2 ([@B74]). CD28 stimulation is also required for the conversion of naïve CD4 T cells into Tregs *in vivo* ([@B75], [@B76]) and *in vitro* ([@B77]).
CTLA-4 in Regulation of the Immune Response {#S4}
===========================================
CTLA-4 is a critical negative regulator of the immune response. Germline CTLA-4 knockout in mice results in massive lymphoproliferation ([@B78]), and is lethal at 3--4 weeks of age ([@B78], [@B79]). The peripheral T cell profile in these animals is strongly skewed toward CD4 cells that rapidly proliferate in a CTLA-4-Ig-sensitive manner -- indicating the dependence on B7--CD28 interaction -- and infiltrate non-lymphoid tissues ([@B78], [@B80]). Introduction of CTLA-4-sufficient Tregs reverts the lymphoproliferative disorder and prevents early lethality in CTLA-4 knockout mice ([@B81]), whereas blocking of CTLA-4 on Tregs completely abrogates their suppressive function ([@B62], [@B66], [@B82]). CTLA-4-deficient Tregs are unable to control lymphopenia-driven homeostatic expansion of conventional CD4 cells ([@B83]). Importantly, interaction between CTLA-4 and B7 expressed on effector T cells was found to be dispensable for the control of the latter in mixed bone marrow chimera experiments as both B7^−/−^CTLA-4^−/−^ and B7^+/+^CTLA-4^−/−^ effector T cells were efficiently suppressed by CTLA-4-sufficient Tregs ([@B66]). B7 expression is also not required on Tregs themselves ([@B66]). These data indicate that B7 expressed on a cell subset distinct from effector and Tregs mediates interactions with Treg-expressed CTLA-4 and immune suppression. CTLA-4-deficient Tregs are characterized by similar expression of CD25, PD1, GITR, and of suppressive cytokines IL-10 and IL-35 ([@B83]). Foxp3 promoter-controlled deletion of CTLA-4 in Tregs resulted in lymphoproliferative disease and tissue infiltration, and was lethal at \~7--8 weeks of age \[i.e., somewhat delayed compared to germline knockout Ref. ([@B84])\]. Similarly to Foxp3-driven CD28 deficiency, thymic development of Tregs was normal, as was their survival in the periphery. However, cells lacking CTLA-4 were unable to control proliferation of target cells stimulated by anti-CD3 antibody and DC, and to induce tumor rejection ([@B84]).
Unlike CD28, CTLA-4 is not required for Treg development in the thymus. CTLA-4 is expressed by a subset of thymocytes predominantly residing at the corticomedullary junction ([@B85]) and is strongly upregulated upon induction of negative selection ([@B86]). There is no requirement for CTLA-4 expression to initiate central Treg development and peripheral expansion, as CTLA-4 knockout mice exhibit elevated percentage of Tregs and increased Ki67 expression, indicative of their active proliferation ([@B87]). Moreover, deletion of CTLA-4 in TCR-transgenic mice increases the frequency of Foxp3-positive Treg precursors in the thymus and leads to the formation of a specific population of Foxp3-positive DP thymocyte subsets in the thymic cortex ([@B85]). However, CTLA-4 can play a role in formation of the induced Treg population, as CTLA-4 has been shown to induce expression of Foxp3 and Treg conversion in the intestine ([@B88]).
CD28 and Immunological Synapse Architecture in Effector T Cells {#S5}
===============================================================
CD28 shows a unique cSMAC localization pattern that is important for its efficient co-stimulatory functions. CD28 co-localizes with TCR microclusters at the earliest observable time-point after agonist pMHC recognition ([@B89], [@B90]), and the early accumulation of CD28 at the IS shows similar kinetics and localization as the TCR complex. In a mature IS, CD28 is present at the cSMAC, but segregates away from TCR ([@B90], [@B91]). This segregation of CD28 from TCR at the IS is required for optimum T cell activation, as shown in a study comparing different anti-CD3 and CD28 micropatterns on planar stimulatory surfaces ([@B92]). The spatial separation of TCR and CD28 at the mature IS is regulated by localization of CD28 ligands, as full length CD80 separates from TCR at the IS, but CD80 with deleted cytoplasmic region localizes with TCR ([@B93]). Moreover, the tailless CD80 molecule does not provide an optimum co-stimulatory signal and does not show efficient accumulation at APC: T cell contact interface ([@B94], [@B95]). This suggests a role for B7 interactions with cytoskeleton and/or other cytoplasmic components in regulation of IS architecture. CD28 recruitment and maintenance at the synapse requires both CD28 and TCR ligand binding ([@B90], [@B96]). CD28 accumulation at the synapse has been shown to be independent of antigenic pMHC affinity to TCR, with weak and strong agonist pMHC complexes inducing similar levels of CD28 recruitment ([@B97]). The role of CD28-mediated signaling and interactions in regulation of CD28 localization at the synapse is somewhat controversial, with a report indicating unperturbed IS localization of CD28 with mutated or deleted cytoplasmic region ([@B90]), whereas another study observed impaired IS localization of CD28 with deleted cytoplasmic domain or with a mutation at Y188 within the CD28 PYAP motif ([@B96]).
CD28 ligation has been shown to induce rapid internalization of the receptor, with half of the endocytosed fraction degraded in lysosomes and half recycled back to the cell surface ([@B98]). CD28 downregulation depends on PI3K ([@B73]), with preferential endocytosis of CD28 molecules associated with PI3K ([@B98]). CD28 is endocytosed via clathrin-coated pits, and this process requires coupling of WASP to PI3K and CD28 via sorting nexin 9 ([@B73]). CD28 downregulation from the synapse can also be influenced by stoichiometry of its B7 ligands ([@B11]). FRET analysis of B7 fluorescent protein fusions demonstrated that CD80 is present at the cell surface as a mixed population of dimers and monomers, with CD86 predominantly present in monomeric form ([@B10]). Experimental increase in CD80 dimerization resulted in enhanced T cell: APC conjugate formation and more sustained accumulation of Lck and PKC-θ at the IS ([@B11]).
Co-stimulatory signals play a critical role in regulation of cytoskeleton dynamics at the IS during T cell interaction with APC (summarized in Table [2](#T2){ref-type="table"}). CD28 ligation induces movement of actin cytoskeleton toward the IS ([@B99]), and CD28 engagement is required for sustained actin accumulation at the IS ([@B100]). CD28 stimulation alone leads to actin polymerization and recruitment of actin at the IS ([@B101]). CD28 signaling is important in multiple pathways involved in actin filament nucleation, elongation, and depolymerization. The guanine nucleotide exchange factor Vav1 controls the activity of small Rho GTPases Cdc42 and Rac1 that regulate actin polymerization activity of WASP and WAVE2, respectively. WASP and WAVE2 are actin nucleation-promoting factors that, together with the Arp2/3 complex, facilitate formation of new actin filaments.
######
**CD28 in regulation of cytoskeleton dynamics at the immunological synapse**.
Cytoskeletal regulator CD28-induced modification Effect on cytoskeleton Reference
------------------------ ----------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------
Vav1 Phosphorylation, leading to activation Vav1 controls activity of small Rho GTPases Cdc42 and Rac1 that regulate actin polymerization activity of WASP and WAVE2, respectively Nunes et al. ([@B102]), Raab et al. ([@B104]), Salazar-Fontana et al. ([@B103]), Schneider and Rudd ([@B106])
Filamin A Direct interaction with CD28, phosphorylation Filamin A has a role in actin crosslinking Muscolini et al. ([@B111]), Tavano et al. ([@B108])
Cofilin Dephosphorylation, leading to activation Actin severing protein. Blocking cofilin--actin interaction reduces T cell:APC conjugation Lee et al. ([@B119]), Wabnitz et al. ([@B120])
Rltpr Unknown Actin-uncapping protein. Wild-type Rltpr is required for CD28-dependent costimulation, but this seems to be independent of its actin-uncapping function Liang et al. ([@B113])
CapZIP Phosphorylation Actin-uncapping protein. CapZIP is required for CD28-dependent costimulation, but its effect on T cell cytoskeleton are unknown Tian et al. ([@B110])
CD28 ligation induces tyrosine phosphorylation of Vav1 ([@B102]), and CD28-dependent actin remodeling requires Cdc42 ([@B103]) and Rac1 ([@B104]). The molecular interactions linking CD28 to Vav1 phosphorylation are not yet fully elucidated. CD28-dependent Vav1 phosphorylation has been shown to require binding of the adaptor protein GRB2 to CD28 ([@B105], [@B106]), but a recent report provided evidence for GRB2-independent Vav1 binding to CD28 and a role of PIP5K1A (phosphatidylinositol 4-Phosphate 5-Kinase α) and Vav1 cooperation in regulation of actin, downstream of CD28 ([@B107]). Jurkat cells expressing CD28 with mutated C-terminal PYAP motif, important for GRB2 binding, failed to recruit Vav1 to the IS or to rearrange actin after CD28 ligation ([@B107]); however, Vav1 phosphorylation in response to CD28 ligation was not assessed in this study, and in another report the PYAP motif was shown to be dispensable for CD28-dependent Vav1 phosphorylation ([@B108]). The Arp2/3 actin nucleation complex cooperates with filamins, large multidomain proteins with a role in actin crosslinking ([@B109]), to establish actin structure. Filamin A is phosphorylated ([@B110]) and recruited to the IS ([@B108]) after CD28 ligation in T cells, with the PYAP motif on the cytoplasmic region of CD28 mediating the interaction with filamin A ([@B108], [@B111]). Knockdown of filamin A expression did not affect CD28-mediated Vav1 phosphorylation, but reduced Cdc42 activity and impaired CD28-mediated costimulation ([@B108]). However, changes in actin structure or dynamics at the IS of filamin A knockdown cells have not been reported. Moreover, knocking down filamin A did not impair ezrin accumulation at the IS ([@B108]).
Actin filaments contain a fast growing barbed end, which can be bound to actin capping protein. Capping protein binding to the barbed end prevents addition of new actin subunits, limiting the filament elongation. Actin capping and subsequent actin polymerization can be regulated by actin-uncapping proteins ([@B112]). An actin-uncapping protein Rltpr is required for CD28-dependent costimulation ([@B113]), and Rltpr colocalizes with CD28 in CD80-dependent signaling microclusters ([@B113]), suggesting a role of Rltpr in CD28-mediated actin rearrangement at the synapse. However, a direct role of Rltpr for CD28-dependent actin modification is unclear. Rltpr does not immunoprecipitate with CD28 ([@B113]) and is not phosphorylated after CD28 ligation ([@B110]). Moreover, an Rltpr mutation that abolishes CD28-mediated costimulation does not impair Rltpr's actin-uncapping ability or CD28-dependent actin rearrangements at the synapse ([@B113]). The Rltpr mutation that reduced CD28-dependent costimulation abrogates CD28-dependent recruitment of PKC-θ and Carma1 to the IS ([@B113]) through a yet unidentified molecular mechanism, suggesting that Rltpr acts as an adaptor at the IS independently of its actin-uncapping functions. A recent phosphoproteomic screen identified actin-uncapping CapZIP as part CD28 signaling network ([@B110]). Importantly, CapZIP is required for CD28-dependent costimulation of cytokine production ([@B110]). However, it has not yet been reported if CapZIP can directly interact with CD28 and if it is required for CD28-dependent changes in actin dynamics. In summary, the current evidence suggests that CD28-dependent signaling may regulate actin capping through actin-uncapping proteins CapZIP and potentially Rltpr.
CD28 signaling regulates actin dynamics through control of activity of the actin-severing protein cofilin. Cofilin is a ubiquitously expressed 19 kDa protein that cleaves actin filaments, thus, promoting actin depolymerization, but also creating new barbed ends for filament elongation ([@B114]). Cofilin's actin binding capacity is negatively regulated by its phosphorylation at serine 3 ([@B115], [@B116]), and binding to phospholipids ([@B117]). Blocking cofilin interaction with actin reduces T cell proliferation and cytokine production, as well as conjugation with APCs ([@B118]). In resting human T cells, cofilin is present mainly in the inactive phosphorylated form, and CD28 or CD2 signal together with TCR, but not TCR signal alone, induces cofilin dephosphorylation and actin binding ([@B119], [@B120]). The precise sequence of signaling events linking CD28 ligation to cofilin activation is unknown. Cofilin is dephosphorylated by serine phosphatases PP1 and PP2A ([@B121]), and CD3/CD28-induced cofilin dephosphorylation requires Ras ([@B120]). Additionally, CD28 may regulate cofilin activity through control of levels of membrane phospholipids ([@B114]).
There is strong evidence that CD28-dependent regulation of actin dynamics is important for the effector T cell functions. CD28 enhances T cell:APC conjugate formation *in vitro* ([@B122], [@B123]). Knock-in mice with mutated PYAP motif show reduced IL-2 production and proliferation *in vitro*, and impaired *in vivo* antibody production and germinal center formation ([@B57]). This could be a result of impaired cytoskeletal rearrangement, as the PYAP motif is implicated in Vav1 and filamin A recruitment. However, the effects of PYAP mutations on cytoskeletal dynamics and synapse stability have not yet been reported for primary T cells, and this motif is also important for binding to Lck ([@B54]), GRB2/GADS ([@B53], [@B55]), and PKC-θ ([@B124]), as discussed below. Analysis of a mouse mutant with inducible inhibition of Csk, a negative regulator of Src family kinases, strongly suggested that CD28-dependent actin remodeling is critical for initiation of full TCR signal in thymocytes ([@B125]). However, thymocytes from PYAP mutant knock-in mice do not show obvious phenotypic defects ([@B57]), suggesting that CD28-independent pathways can regulate actin cytoskeleton dynamics during thymocyte development.
CD28 and Regulation of PKC-θ Localization at the Effector T Cell IS {#S6}
===================================================================
CD28 plays a critical role in regulation of the IS localization of the novel protein kinase C (nPKC) isoform PKC-θ (summarized in Table [3](#T3){ref-type="table"}). The PKC family consists of 10 serine/threonine kinase isoforms, with important roles in regulation of multiple cellular processes in different cell types. All nPKC isoforms (PKC-θ, PKC-δ, PKC-ϵ, and PKC-η) require diacylglycerol (DAG), but not Ca^2+^, for activation, and are expressed in T cells and play multiple roles in regulation of T cell signaling and effector functions ([@B126]). Central localization of PKC-θ is one of the hallmarks of the mature effector T cell IS. A seminal study by Monks at al. identified PKC-θ as the only PKC isoform recruited to effector T cell IS ([@B127]). However, more recent studies show that PKC-η and PKC-ϵ are also recruited ([@B128]--[@B130]), with some evidence that their recruitment precedes that of PKC-θ ([@B129]). PKC-ϵ and PKC-η display homogeneous distribution over the entire synapse, whereas PKC-θ displays discrete cSMAC localization contained within the peripheral actin ring ([@B128]--[@B132]).
######
**Molecular determinants of PKC-θ localization at the immunological synapse**.
Interaction/activity Molecular determinants Effect on immunological synapse Reference
----------------------- ----------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------
PKC-θ--CD28 Polyproline motif within the PKC-θ V3 hinge region and PYAP motif in CD28; Lck-mediated interaction PKC-θ V3 hinge and CD28 PYAP motif are required for CD28 cSMAC localization Kong et al. ([@B124])
PKC-θ--CD28 Sumoylation of PKC-θ at lysines 325 and 506 Abrogated PKC-θ sumoylation reduces PKC-θ localization at the IS and its colocalization with CD28, induces colocalization of PKC-θ and filamin A at periphery of the IS Wang et al. ([@B139])
PKC-θ--DAG C1 domains of PKC-θ C1 domains mediate initial PKC-θ recruitment to the synaptic membrane, but they do not support PKC-θ central accumulation at the synapse Basu et al. ([@B134]), Carrasco and Merida ([@B136]), Quann et al. ([@B135])
PKC-θ kinase activity Unknown, possibly through autophosphorylation at threonine 219 between the tandem C1 domains PKC-θ kinase activity is required for its recruitment to the IS Cartwright et al. ([@B138]), Thuille et al. ([@B137])
Rltpr Unknown, no interaction between Rltpr and PKC-θ has been detected Wild-type Rltpr is required for PKC-θ and CARMA1 recruitment to cSMAC Liang et al. ([@B113])
An important study using lipid bilayers presenting antigen and co-stimulatory signal, and TIRF microscopy to examine PKC-θ localization at the effector T cell IS, revealed initial colocalization of PKC-θ with TCR/CD28 microclusters ([@B90]). This was followed by PKC-θ recruitment to the cSMAC, where it segregated, together with CD28, to TCR^low^ regions in the periphery of cSMAC ([@B90]). The initial stages of PKC-θ recruitment to the effector T cell IS do not depend on CD28 ligand binding, but CD28 ligation is required for sustained PKC-θ localization at the synapse and colocalization of PKC-θ with CD28 ([@B90], [@B133]). PKC-θ interacts with CD28 after PMA treatment ([@B90]) (which induces PKC activation) and TCR/CD28 stimulation ([@B124]).
The molecular determinants of PKC-θ synapse localization have been mapped to the V3 hinge region and C1 domains ([@B132]). nPKCs share a conserved structure, with an amino-terminal C2 domain, tandem C1 domains, and V3 hinge linked to a carboxyl-terminal kinase domain. PKC-θ interaction with CD28 and cSMAC localization requires a polyproline motif within the V3 hinge region ([@B124]), and V3 hinge regions from PKC-ϵ and PKC-η mediate their diffuse accumulation at the synapse ([@B134]). A carboxyl-terminal poly-proline motif (PYAP) in the CD28 cytoplasmic tail is required for its interaction with PKC-θ, with strong evidence suggesting that this is an indirect interaction mediated through Lck, with the Lck SH3 domain binding to the polyproline motif in PKC-θ V3 and the Lck SH2 domain binding a phosphorylated tyrosine within the CD28 PYAP motif ([@B124]). Tyrosine 188 within the PYAP motif was also identified as critical for CD28 and PKC-θ central synapse localization in an earlier study ([@B96]).
Additionally, C1 domains of PKC-θ also play a role in its synapse localization, through interaction with DAG at the synapse membrane ([@B134], [@B135]). C1 domains can mediate initial PKC-θ recruitment to the synaptic membrane ([@B135]), but they do not support PKC-θ central accumulation and retention and the membrane ([@B136]), and nPKC C1 domains are not sufficient to determine the respective synapse localizations of PKC-θ versus PKC-ϵ and PKC-η ([@B134]). Phosphorylation of PKC-θ threonine 219 (T219), in a hinge region between the tandem C1 domains, is required for PKC-θ localization at the IS ([@B137]). Moreover, sustained synapse localization is dependent on PKC-θ kinase activity ([@B137], [@B138]), most likely through a requirement for PKC-θ autophosphorylation at T219 ([@B132], [@B137]). PKC-θ recruitment to the IS also requires expression of wild-type Rltpr actin-uncapping protein ([@B113]). T cells from mice expressing an Rltpr mutant could not recruit PKC-θ to the IS ([@B113]). The precise role of Rltpr in the regulation of PKC-θ synapse localization is unknown but seems to be independent of Rltpr actin-uncapping function, and no direct interactions between Rltpr and PKC-θ have been observed ([@B113]).
A recent report identified a novel activation-dependent post-translational modification of PKC-θ that modulates CD28--PKC-θ interactions and IS architecture ([@B139]). TCR stimulation of resting murine and human T cells leads to conjugation of SUMO1 (small ubiquitin-like modifier) to PKC-θ lysine (K) 325 and K506 by SUMO E3 ligase PIASxβ ([@B139]). Importantly, TCR and CD28 costimulation resulted in stronger PKC-θ sumoylation than TCR stimulation alone. Sumoylation-resistant PKC-θ with mutated K325 and K506 residues showed reduced interaction with CD28 and filamin A, and diffuse localization at the membrane in the IS ([@B139]). Inhibiting PKC-θ sumoylation through PIASxβ knockdown or overexpression of a desumoylating enzyme also abrogated PKC-θ localization at the IS, and reduced its colocalization with CD28 ([@B139]). Wild-type PKC-θ segregated from filamin A at the IS, with mainly pSMAC localization of filamin A. Inhibition of PKC-θ sumoylation altered the IS architecture, inducing colocalization of PKC-θ and filamin A at the periphery of the synapse ([@B139]).
The localization of PKC-θ to the center of the IS is critical for its functions in effector T cells. Mutations of the poly-proline motif in the V3 region of PKC-θ reduced activation of primary effector CD4+ T cells ([@B124]). Critically, overexpression of murine V3 domain sequesters PKC-θ from CD28 and cSMAC in mouse CD4+ T cells, and reduces PKC-θ-dependent gene expression *in vitro*, as well as CD4+ Th2 and Th17 immune responses *in vivo* ([@B124]). Similarly, expression of sumoylation-resistant PKC-θ mutants, with impaired synapse localization, does not rescue defects in cytokine production, activation of PKC-θ dependent transcription factors, and Th2 differentiation of human T cells with downregulated expression levels of endogenous PKC-θ ([@B139]). Additionally, mutations in the CD28 PYAP motif, required for PKC-θ interaction with CD28 and for IS localization, severely impaired effector T cell functions *in vivo* ([@B57]). However, it must be noted that PKC-θ synapse localization seems to be inseparable from its interaction with CD28, and the observed functional effects of impaired PKC-θ synapse recruitment could also be caused by reduced CD28 interactions with PKC-θ, Lck, and/or filamin A.
CTLA-4 Dynamics at the Effector and Treg IS {#S7}
===========================================
Recognition of B7 ligands by CD28 and CTLA-4 at the effector T cell IS leads to competitive displacement of CD28 and PKC-θ from its central region. In the absence of stimulation, regulatory and conventional T cells express similar levels of CD28, but CTLA-4 expression is significantly higher in unstimulated Tregs ([@B71], [@B140], [@B141]). TCR signaling induces polarization of both intracellular ([@B142]) and membrane pools of CTLA-4 toward the IS of effector T cells, and TCR signal strength determines CTLA-4 localization at the IS ([@B97]). CTLA-4 is recruited to the effector T cell cSMAC with delayed kinetics relative to that of TCR and CD28, segregates away from CD3^high^ regions and forms a ring-like structure ([@B141]). CTLA-4 recruitment to and stabilization at the IS depends on its ligand binding, but occurs at both high and low B7 ligand densities ([@B141]). Critically, recruitment of CTLA-4 to the IS influences CD28 localization, due to competition for ligand binding. At high ligand densities, CTLA-4 recruitment leads to exclusion of CD28 from the cSMAC and its accumulation outside the pSMAC ([@B141]). At low ligand densities, CTLA-4 prevents formation of CD28 clusters at the T effector IS ([@B141]). Importantly, CTLA-4-mediated displacement of CD28 from the cSMAC leads to impaired synaptic localization of PKC-θ ([@B141]). CTLA-4 ligation has also been reported to reduce the size of T cell: APC contact interface and to reduce ZAP70 microcluster formation ([@B143]).
The localization of CTLA-4 in the T effector synapse depends on the molecular dimensions of the extracellular region of the protein, as CTLA-4 molecules with elongated ectodomains failed to accumulate at cSMAC despite unimpaired ligand binding ([@B141]). However, it has not been reported if CTLA-4 with elongated ectodomains affected CD28 clustering at the synapse, and it is not known whether similar dimensions of CD28 and CTLA-4 receptor--ligand complexes are important for efficient regulation of co-stimulatory signal and/or competition for ligand binding at the synapse. The matching sizes of activating and inhibitory receptor--ligand complexes are critical for signal integration and regulation of NK-cell functions ([@B144], [@B145]), and it is plausible that a similar requirement exists for co-stimulatory and co-inhibitory signaling in effector T cells.
A molecular mechanism for CTLA-4 involvement in the downregulation of CD80/CD86 has been established in the seminal work by Qureshi et al. ([@B146]). Using co-cultured CHO cells expressing either human CTLA-4 or GFP-tagged human CD86, they observed transfer of GFP signal into CTLA-4 expressing cells, and its accumulation in the endolysosomal system, indicative of CD86 trans-endocytosis. Endocytosis-deficient CTLA-4 failed to induce trans-endocytosis of CD86-GFP and resulted in the accumulation of CD86 at cell contacts. These findings were confirmed using purified human Tregs incubated with DC where CD86 expression on the surface of the DC was reduced in the presence of Tregs but not effector T cells, and TCR stimulation increased the rate of trans-endocytosis ([@B146]). The most direct consequence of reduction of B7 proteins on the surface of APC is manifested in fewer and less prolonged interactions between APC and effector T cells ([@B147], [@B148]) reduced PKC-θ recruitment and activation in these cells ([@B149]) and, consequently, repression of IL-2 production by effector T cells ([@B150]). Recently, it has been shown that surface expression of CTLA-4 on effector T cells is sufficient for downregulation of CD86 expression from APCs ([@B40]).
Tregs display radically different synapse localization of CD28 and PKC-θ than effector T cells (summarized in Figure [2](#F2){ref-type="fig"}). In a stimulating planar lipid bilayer system, the recruitment of CD28 to the Treg IS is barely detectable, whereas CTLA-4 is recruited robustly, in stark contrast to conventional CD4 T cells ([@B141]). Displacement of CD28 from Treg synapses by CTLA-4 coincides with the absence of PKC-θ clusters in the cSMAC zone of Treg synapses. Similarly, the switch of developmental program during the *in vitro* conversion of naïve CD4 T cells into Tregs results in a loss of PKC-θ signal at the synapse. Correct localization of CTLA-4 to the IS is functionally important, as elongation of the extracellular domain of CTLA-4 resulted in a loss of its concentration in the synapse and reduction of suppressive activity of Tregs ([@B141]). In a lipid bilayer system, addition of CD80 or ICAM-1 to the bilayer increases the recruitment of PKC-θ to the synapse in both effector and Tregs, but stimulation through TCR strongly decreases the recruitment in Tregs ([@B151]). Reduction of PKC-θ activity results in increased Treg proliferation and elevated suppressive capacity ([@B151]).
{#F2}
An important insight into the signaling mechanism downstream of CTLA-4 recruitment to the Treg synapse was provided in a recent study by Kong et al. ([@B152]) which identified PKC-η as the only PKC isoform physically interacting with CTLA-4. In Tregs, a phosphorylated form of PKC-η binds constitutively to CTLA-4. PKC-η localizes to the IS in close proximity to the TCR ([@B152]). Interaction between PKC-η and CTLA-4 was found to be critical for Treg function. Phosphorylated serine residues S28, S32, and S317 of PKC-η are responsible for the interaction with CTLA-4, and loss of S28 or S32 results in a strong inhibition of some Treg suppressive functions. The importance of PKC-η was further emphasized by the finding that, although PKC-η-deficient Tregs expressed normal levels of functional LFA-1 required for the stabilization of contacts with APC, they showed a marked decrease in their ability to continuously clear CD86 from the APC surface. While CD86 clearance on first contact with APC was unaffected by the loss of PKC-η, the reduction of CD86 on reintroduced APC was substantially delayed ([@B152]). These findings suggest that PKC-η is not directly involved in CTLA-4-induced trans-endocytosis, and that a feedback signaling mechanism from PKC-η may be required for the recycling of CTLA-4 from the endolysosomal system. It remains to be elucidated whether or not the amount of surface-expressed CTLA-4 is reduced and its intracellular retention or lysosomal degradation is accelerated in the absence of PKC-η. In an earlier study, a deletion of amino acids 191--223 of the intracellular domain of CTLA-4 did not substantially affect the *in vitro* suppression of target T cell proliferation in the presence of CD3 crosslinking antibody and APC, or *in vivo* suppression in a colitis model ([@B26]). Kong et al. have shown that this deletion-mutant of CTLA-4 retains its association with PKC-η, suggesting that the remaining cytoplasmic portion (amino acids 182--191) of CTLA-4 is sufficient for suppressive signals using PKC-η.
Responses to CD80 and CD86 signals in effector T cells are largely similar ([@B153], [@B154]). By contrast, addition of anti-CD80 antibody or CTLA-4 Fab fragments to the co-culture of target T cells, DC, and prestimulated Tregs ablated the suppressive function of the latter, whereas addition of anti-CD86 or anti-CD28 antibodies increased suppression to the same degree ([@B155]). Furthermore, blocking CD86 inhibited DC-induced division of Tregs, whereas blocking CD80 enhanced division ([@B155]). These data suggest that, in contrast to effector T cells, Tregs can effectively discriminate between CD80--CTLA-4 and CD86--CD28 signals.
CTLA-4 in Regulation of Regulatory and Effector T Cell Synapse Stability and Cell Motility {#S8}
==========================================================================================
A growing body of evidence clearly suggests a role for CTLA-4 in regulation of synapse stability, duration of conjugation with APC and overall motility in Treg and Teff cells. Anti-CTLA-4 blocking antibody treatment has been shown to increase effector T cell motility *in vitro* ([@B122], [@B156]) and *in vivo* ([@B157]--[@B159]). Importantly, it has been reported that CTLA-4 ligation has different outcomes for synapse stability and motility of regulatory versus effector T cells ([@B122], [@B160]).
Regulatory T cells form a more stable IS than effector T cells, and this enhanced synapse stability has been implicated in CTLA-4-dependent downregulation of B7 cell surface expression by Tregs. In mixtures with conventional CD4 T cells of same specificity, TCR-transgenic Tregs preferentially bind to DCs and exclude conventional cells ([@B161]). Similarly, in a planar lipid bilayer system, Tregs form a more long-lived IS than do effector T cells of the same specificity ([@B151]). Addition of blocking CTLA-4 antibody does not overrule the competitive advantage of Tregs, but loss of LFA-1 results in its reversal, indicating that LFA-1 is at least partially responsible for the preferential binding of Tregs ([@B161]). Stimulated TCR-transgenic Tregs specifically reduce expression of both CD80 and CD86 on DC and to lesser degree B cells ([@B84], [@B87], [@B161], [@B162]). In the absence of stimulating peptide, the B7 molecules are not downregulated. CTLA-4-deficient Tregs as well as wild-type cells in the presence of CTLA-4 blocking antibody do not reduce B7 expression, and neither do Tregs from LFA-1^−/−^ mice ([@B84], [@B161]), indicating that the formation of a stabilized, LFA-1 dependent, Treg-APC IS is important for the B7 downregulation.
B7 ligand recognition can modulate Treg motility. Tregs migrate rapidly on non-stimulating bilayers but slow down significantly, and increase contact time, upon encountering a TCR signal (TCR stop signal) ([@B160], [@B163]). Importantly, the stop signal required for the slowing down of Tregs is CTLA-4-independent, as CTLA-4-deficient TCR-transgenic Tregs slowed down as efficiently as CTLA-4-sufficient cells in mixed culture with antigen-pulsed DC ([@B160]). Similarly, addition of blocking CTLA-4 antibody to Tregs on lipid bilayers containing CD80, pMHC, and ICAM-1 did not affect their motility ([@B163]). Although active displacement of CD28 from the IS by CTLA-4 precludes an active role for the former in the establishment and stabilization of Treg--APC synapses, a growing body of data indicates that CD28 may be important for the orchestration of Treg motility and contact half-life with APC prior to mature synapse formation. However, data on involvement of CD28 in stop signaling remain contradictory. While CD28-deficient Tregs stop normally in mixed culture ([@B160]), CD28-blocking antibody interfered with the stop signal ([@B163]). Specific loss of CD28 in Tregs reduces surface expression of CTLA-4 on these cells and results in reduced suppressive capacity and systemic autoimmunity ([@B71]), indicating that tonic signaling input downstream of transient B7--CD28 interactions may regulate recycling of CTLA-4 protein. CD86 and CD28 input is also important for DC-induced proliferation of Tregs ([@B155]).
Unlike Tregs, effector T cells are sensitive to CTLA-4-dependent reversal of the TCR stop signal. In the first report of CTLA-4 dependent reversal of TCR stop signal, Schneider et al. ([@B156]) used anti-CTLA-4 stimulation and observed that it enhanced effector T cell motility on LFA-1-coated plates. Moreover, anti-CD3 antibody induced reduction in T cell motility, but a combination of anti-CD3 and anti-CTLA-4 did not elicit this stop signal. In the same study, CTLA-4 expression on effector T cells increased their motility and reduced their contact time with APCs in the context of antigen recognition in the lymph node ([@B156]). CTLA-4 was also shown to reverse the TCR stop signal in human effector T cell clones *in vitro* ([@B122]). Additionally, CTLA-4 antibody treatment enhanced effector T cell motility in the context of an anti-tumor response ([@B158], [@B159]). However, two other 2 photon imaging studies did not report this differential effect of CTLA-4 blockade on regulatory and effector CD4+ T cell populations ([@B157], [@B164]). A study using a mouse model of T cell responses in pancreatic islet grafts reported that CTLA-4 blockade slightly increased motility of both effector and Treg populations, suggesting that CTLA-4 marginally reduces CD4+ T cell motility *in vivo* ([@B157]). However, the imaging performed in this study was conducted in an immunoprivileged site (the islet grafts were injected into the anterior chamber of the eye), which could have affected the cellular motility observed. Moreover, the role of TCR signal in the reported effector and Treg motility changes is unknown, as this study observed direct interactions between pancreatic peptide-specific TCR-transgenic CD4 T cells and islet cells, which do not express MHC class II and, thus, cannot present antigenic peptide to T cells. Another study investigating motility of tolerized diabetogenic CD4+ T cells reported no effect of CTLA-4 blockade on T cell motility ([@B164]). However, this study did not differentiate between effector and Treg populations, and the effect on CTLA-4 on control, non-tolerized diabetogenic T cells was not reported ([@B164]). Overall, the results from these two studies indicate that CTLA-4 has limited effect on motility of self-antigen-specific CD4+ effector T cells, similar to its relatively limited effect on Treg motility. Given that the natural Treg lineage consists of self-reactive T cells ([@B165]), this raises an interesting possibility that the role of CTLA-4 in regulation of synapse stability and cellular motility of CD4+ T cells depends on their TCR specificity.
Importantly, the different effects of CTLA-4 blockade on effector T cells and Tregs have also been observed in a recent study using 2 photon microscopy to examine the behavior of the two CD4+ T cell populations in intact lymph nodes ([@B13]). CTLA-4 blockade increased Treg motility but decreased effector T cell motility in the presence of antigen, consistent with the proposed role of CTLA-4 in reversal of TCR-induced motility stop in effector, but not regulatory, T cell populations. Anti-CTLA-4 antibody administration increased effector T cell contact time with DCs presenting antigen, but reduced Treg contacts with DCs, strongly suggesting that CTLA-4 has opposing effects on effector and Treg IS stability *in vivo*. However, as CTLA-4 blockade increased the steady-state motility of Tregs, but had no effect on effector T cell motility in the absence of antigenic stimulation, the reduced effector T cell motility and enhanced clustering with DCs after anti-CTLA-4 treatment could be the result of exclusion of Tregs from T cell: DC clusters, rather than a direct effect of CTLA-4 on effector T cells. Interestingly, this study also reported regulatory-effector T cell contacts that were dependent on Treg recognition of B7 expressed on activated T cells ([@B13]), suggesting that CTLA-4: B7 interaction plays a role in regulation of T:T cell synapse formation and facilitates Treg-mediated suppression.
The molecular mechanism of CTLA-4-dependent regulation of effector T cell synapse stability and cellular motility is unknown. The initial observation that anti-CTLA-4 treatment enhances effector T cell motility on LFA-1 coated slides and in response to TCR signal was originally interpreted as evidence for an as yet unidentified CTLA-4-induced signal overriding the TCR stop signal ([@B156]). CTLA-4 ligation was shown to reduce IS stability ([@B122], [@B166]) and decrease cytoskeletal rearrangements at the synapse ([@B166]) through an unknown molecular mechanism. CTLA-4 ligation was also shown to activate the small G protein Rap1 ([@B167], [@B168]), and CTLA-4-induced Rap1 activity was linked to destabilization of the IS ([@B53]). However, CTLA-4 mediated increase in Rap1 activity has also been linked to enhanced LFA-1 mediated adhesion ([@B167]--[@B169]), which is difficult to reconcile with the reduced synapse stability. Moreover, an *in vivo* study reported that intact anti-CTLA-4 antibody and its Fab fragments enhanced effector T cell motility equally well, suggesting that CTLA-4-dependent signaling did not play a role in the motility enhancement ([@B158]). Given the role of CD28 in regulation of cytoskeletal dynamics, it is plausible that CTLA-4 may reduce synapse stability and enhance T cell motility through counteracting CD28-mediated cytoskeletal rearrangement through competition for B7 ligand binding. However, there is conflicting evidence to support this hypothesis. It has been reported that the CTLA-4-dependent increase in motility does not require CD28 expression ([@B160]), and that the cytoplasmic region of CTLA-4 is required for regulation of T cell motility ([@B166]), suggesting a role for as yet unidentified CTLA-4-dependent signaling.
Conclusion {#S9}
==========
B7 ligand recognition plays an important role in orchestrating the IS architecture in both effector T cells and Tregs. During recognition of antigen, B7 ligand binding induces CD28 localization to distinct TCR^low^ clusters within the central region of the effector T cell IS. This CD28 recruitment can be counteracted by CTLA-4 through competition for ligand binding and/or by removal of co-stimulatory ligands through trans-endocytosis. CD28 recruitment to the IS induces PKC-θ localization to the center of the IS, through interactions between PKC-θ V3 hinge region and the proline-rich motif on the cytoplasmic tail of CD28. CD28 ligation leads to cytoskeletal rearrangements at the IS through CD28-dependent control of multiple pathways regulating cytoskeletal dynamics: Vav1 and cofilin activation, filamin A binding, and regulation of actin-uncapping proteins. CD28-dependent PKC-θ recruitment and modulation of cytoskeleton plays a critical role in regulation of effector T cell functions. CTLA-4 is a negative regulator of effector T cell functions, and there is evidence that CTLA-4 can reduce effector T cell IS stability through reversal of the TCR-induced stop signal.
While signaling through CD28 is important for steady-state homeostasis, motility, target recognition, and division of Tregs, their activation results in active exclusion of CD28 and PKC-θ and recruitment of CTLA-4 and PKC-η to the synapse. Both phenomena are required for the suppressive function. In contrast to conventional CD4 T cells, CTLA-4 and PKC-η act as positive regulators of Treg function. Since CTLA-4 binds to B7 proteins with significantly higher affinity than CD28 and exclusively activates PKC-η ([@B152]), it is reasonable to conclude that its affinity alone may be sufficient to initiate the exclusion of CD28 from potential B7 binding sites. Preferential activation of PKC-η is then a direct outcome of exclusive CTLA-4 recruitment. Higher affinity of CTLA-4--B7 interactions also explains why Tregs are capable of actively recruiting B7 proteins to the synapse, while effector T cells are not ([@B170]). Increased affinity of B7--receptor interaction and recruitment of B7 proteins to the synapse also contribute to more long-lived, stable synapses between Tregs and APCs as compared to conventional T cells.
B7 ligand recognition induces dissimilar immune synapse architecture in mature effector T cells and Tregs, but its role in regulation of the immune synapse dynamics at different stages of T cell development is poorly understood. It remains to be determined if the IS formed by pre-selection thymocytes shows the central localization of CD28 and/or PKC-θ, similar to that observed in effector T cells. Given that immature thymocytes do not show the centralized TCR accumulation at the immune synapse ([@B171], [@B172]), and that PKC-θ is not required for NFκB activation in thymocytes ([@B173]), it is likely that CD28/PKC-θ dynamics at the thymocyte IS are different to the dynamics in mature effector T cells. At the other end of the T cell's lifetime, CD28/PKC-θ immune synapse dynamics in exhausted T cells or T cells from aged individuals are very poorly understood. Anergic murine T cells were shown to display unimpaired PKC-θ recruitment ([@B174]), but loss of CD28 from human T cells due to repeated antigen exposure or aging may have implications on PKC-θ synapse localization, resulting in altered kinetics and architecture of the synapse, and changes in downstream signaling. Furthermore, despite the massive amount of data on CTLA-4 biology and the growing importance of the CTLA-4 pathway in immunotherapy, its role in regulation of T cell functions and IS dynamics remains incompletely understood. Since activated effector T cells express CTLA-4 and since surface CTLA-4 is capable of B7 extraction from target membranes regardless of cell type ([@B146]), it will be intriguing further to explore the potential role of CTLA-4 in the effector T cell-intrinsic restriction of strength and/or duration of activation, independently of bystander suppression by Treg-expressed CTLA-4. At the signaling level, given the role of CD28 signaling in regulating cytoskeleton dynamics at the IS, CTLA-4 can likely counteract these CD28-mediated pathways, either indirectly through reducing levels of B7 proteins on APCs, or directly through interactions yet unidentified with other binding partners. PKC-η is a likely candidate, as Tregs lacking PKC-η showed enhanced conjugation with DCs, and the CTLA-4--PKC-η complex has been shown to interact with the focal adhesion complex components PAK2 and GIT2, as well as with the guanine nucleotide exchange factor αPIX ([@B152]), with a known role in regulating cytoskeletal dynamics ([@B175]). Better understanding of the effect of B7 ligand recognition on the IS dynamics at different stages of T cell development and in different T cell subsets is likely to have significant implications for the development of novel immunotherapy strategies.
Author Contributions {#S10}
====================
VR and JB wrote the initial draft of this review. VR, NG, and JB rewrote and edited the article.
Conflict of Interest Statement {#S11}
==============================
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.
Funding {#S12}
=======
This work was funded by grants from the Ministry of Education (MOE2014-T2-1-136) and National Medical Research Foundation (CBRG0064/2014).
[^1]: Edited by: Andres Alcover, Institut Pasteur, France
[^2]: Reviewed by: Janis K. Burkhardt, Children's Hospital of Philadelphia, USA; Bernard Malissen, Centre d'Immunologie de Marseille Luminy, France
[^3]: Specialty section: This article was submitted to T Cell Biology, a section of the journal Frontiers in Immunology
| {
"pile_set_name": "PubMed Central"
} |
###### Summary box
What is already known about this subject?
=========================================
- The 2019 new coronavirus pneumonia is a new infectious clinical disease caused by a new coronavirus, called SARS-CoV-2.
- Fever and cough were the most common symptoms in patients with COVID-19, but some non-classical symptoms were underestimated.
What are the new findings?
==========================
- Gastrointestinal (GI) symptoms accounted for a certain proportion in patients with COVID-19, though fever and cough still were the main symptoms.
- There was no statistically significant difference in severity of COVID-19 between patients with and without GI symptoms, but the proportion of severe disease in patients with GI symptoms was higher than in patients without GI symptoms.
How might it impact on clinical practice in the foreseeable future?
===================================================================
- In clinical practice, the presence of GI symptoms in patients does not indicate a risk of disease progression, which cannot be a predictor.
Introduction {#s1}
============
Since December 2019, novel coronavirus (SARS-CoV-2)-infected disease (COVID-19) has gradually swept the world. Morbidity and mortality are increasing due to the high infectivity of the disease worldwide. By 20 April 2020, a total of 2 291 281 infections and 160 044 deaths had been confirmed in 211 countries and regions, with a case fatality rate of 7%, and the number of infections and deaths was dramatically rising daily. SARS-CoV-2 belongs to the same coronavirus family as SARS-CoV and MERS-CoV.[@R1] But its transmission speed is higher than the other two, and transmission channels are more diversified, including respiratory and digestive tract.[@R2] Earlier studies indicated that the clinical symptoms of COVID-19 vary and not all appear as respiratory symptoms. In some cases, the main symptoms are gastrointestinal (GI) symptoms such as abdominal pain, diarrhoea, nausea and vomiting.[@R4] Moreover, there are many patients of severe disease with GI symptoms.[@R5] Hence, it is essential to explore the correlation between digestive symptoms and disease severity, so as to pre-estimate the disease severity and give appropriate early special care and treatment.
Material and methods {#s2}
====================
Search strategy {#s2-1}
---------------
Six medical databases were searched including three English databases (Pubmed, Cochrane Library and Embase) and three Chinese databases (CNKI, Wan Fang Data, China Science and Technology Journal Database). The keywords are as follows: 'COVID-19', 'gastrointestinal symptoms' and so on.
Inclusion/exclusion criteria {#s2-2}
----------------------------
Inclusive criteria: (1) research types : cross-sectional studies, case control studies and case series; (2) research subjects: patients with confirmed COVID-19; (3) data items: including demographic characteristics (age and gender), clinical characteristics (fever, dry cough, fatigue, nausea and vomiting, abdominal pain and diarrhoea) and comorbidities (hypertension and diabetes). Exclusive criteria: (1) the type of study is case report, review and so on; (2) repeated research; (3) lack of the above case data; (4) animal experiments.
Data extraction and paper quality evaluation {#s2-3}
--------------------------------------------
A total of 2324 articles were retrieved. Browse the titles and abstracts, remove duplicate references, leaving 57 studies. After reading the full text, 48 articles were removed. Finally, a total of 9 studies[@R5] (including 8 English articles and 1 Chinese literature) and 3022 patients were included ([figure 1](#F1){ref-type="fig"}). The Newcastle-Ottawa Scale (NOS) scoring system scored an average of 7.7 (6--8), indicating that the quality of selected studies was relatively high.
{#F1}
Statistical analysis {#s2-4}
--------------------
Meta-analysis was performed using Revman5.3. I² statistics are calculated to measure the proportion of total variation in study estimates attributed to heterogeneity. The combined OR and 95% CI (p\<0.05) were calculated for heterogeneity. The forest plot and funnel plot were developed to assess heterogeneity and publication bias.
Result {#s3}
======
Demographical characteristics, clinical features and comorbidities {#s3-1}
------------------------------------------------------------------
The main characteristics of all included studies are as follows: 50% patients are male (95% CI 0.482 to 0.518) and mean age of patients was 49.7 (95% CI 49.213 to 50.244) ([figure 2](#F2){ref-type="fig"}). A total of 479 patients with COVID-19 (13.7%, 95% CI 0.125 to 0.149) had severe disease ([figure 3](#F3){ref-type="fig"}) and 624 patients (14.7%, 95% CI 0.136 to 0.159) had GI symptoms. The incidence of fever was 85.9% (95% CI 0.846 to 0.872), that of dry cough was 60.5% (95% CI 0.588 to 0.623) and that of fatigue was 32.2% (95% CI 0.305 to 0.339) ([figure 4](#F4){ref-type="fig"}). The incidence of comorbidities, including hypertension and diabetes was 17.1% (95% CI 0.156 to 0.185) and 8.1% (95% CI 0.070 to 0.091), respectively ([figure 5](#F5){ref-type="fig"}).
{#F2}
{#F3}
{#F4}
{#F5}
Gastrointestinal symptoms {#s3-2}
-------------------------
As shown in [table 1](#T1){ref-type="table"}, of 624 patients with COVID-19 with GI symptoms, 118 patients had severe disease (20.5%, 95% CI 0.133 to 0.276) and of 2397 cases without GI symptoms, 361 patients were severe (18.2%, 95% CI 0.129 to 0.235) ([figure 6](#F6){ref-type="fig"}). Comparing the severity of COVID-19 in patients with and without GI symptoms, the results indicated: I²=62%; the heterogeneity among studies was moderate. For OR=1.21, 95% CI 0.94 to 1.56, and p=0.13, there was no statistically significant difference in severity of COVID-19 between patients with and without GI symptoms ([figure 7](#F7){ref-type="fig"}). The funnel plot was symmetrical with no publication bias([figure 8](#F8){ref-type="fig"}). The outcome showed that there was no significant correlation between GI symptoms and disease severity in patients with COVID-19.
######
Severe patients with and without GI symptoms
GI symptoms No GI symptoms
--------------------- ---------------- ----- ----- ------
Zhou *et al*[@R6] 4 66 12 188
Jin *et al*[@R5] 17 74 47 577
Guan *et al*[@R7] 22 97 151 1002
Zhang *et al*[@R8] 24 55 34 84
Zhou *et al*[@R9] 5 16 49 175
Fang *et al*[@R10] 17 159 29 146
Lin *et al*[@R11] 14 58 6 37
Nobel *et al*[@R12] 14 97 30 181
Chen *et al*[@R13] 1 2 3 7
Total 118 624 361 2397
GI, gastrointestinal.
{#F6}
{#F7}
{#F8}
Discussion {#s4}
==========
In this study, fever (85.9%) and respiratory symptoms (60.5%) were still the main manifestations in patients with COVID-19, but GI symptoms (14.7%), such as nausea, vomiting, diarrhoea and abdominal pain, also appeared in a considerable number of patients, which was consistent with the previous study that Huang *et al*[@R14] conducted. Wang *et al*[@R15] reported that abdominal pain was more frequent in patients who required intensive care unit care than those who did not. In addition, Holshue *et al*[@R16] and Gui *et al*[@R17] suggested that the crucial ACE2 receptor of SARS-CoV-2 infecting cells was expressed in lung AT2 cells and in intestinal epithelial cells. The study conducted by Lukassen *et al* showed that organs with ACE2 expressing cells may be considered as potential infection sites and transmission routes for SARS-CoV-2.[@R18] Liang *et al*[@R19] reported that due to the high ACE2 expression in proximal and distal enterocytes, intestine may be vulnerable to SARS-CoV-2 infection. GI symptoms, such as vomiting and diarrhoea, lead to the interruption of intestinal flora and electrolyte disturbance such as low potassium and imbalance of water and sodium. This is likely to worsen the patients' condition. However, the result of this meta-analysis showed that there was no statistical correlation between the presence of GI symptoms and the severity of COVID-19. Although there was no statistical significance, the result that the proportion of severe disease in patients with GI symptoms (20.5%) was higher than that in patients without GI symptoms (18.2%) was evident. This study also has some limitations: (1) the languages of retrieval literature are limited to Chinese and English and all the articles are limited to published literature and (2) GI symptoms may be under-reported in some studies, which may cause a lower pooled prevalence rate. Despite the limitations, this meta-analysis overcomes the shortcomings of small sample size and regional restrictions. The heterogeneity and publication bias among the studies are moderate and the results are relatively objective.
In summary, current findings are not sufficient to demonstrate a significant correlation between GI symptoms and disease severity in patients with COVID-19. Large multicentric prospective studies are required to confirm our findings.
The authors would like to thank the Department of Gastroenterology, Xinjiang Medical University Affiliated First Hospital for support of funding this research. The authors would also like to thank the frontline medical staff of Xinjiang Province for their bravery and efforts in SARS-CoV-2 prevention and control.
**Contributors:** JL designed the study, analysed the data and wrote the paper. MC and TY collected data and performed the study. PY designed the study, supervised the whole study process and critically revised the manuscript.
**Funding:** This study was funded by National Science Funding of China (81760100).
**Competing interests:** None declared.
**Patient consent for publication:** Not required.
**Provenance and peer review:** Not commissioned; externally peer reviewed.
**Data availability statement:** All data relevant to the study are included in the article or uploaded as supplementary information.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Ischemic preconditioning of a limb (IPC) is a non-invasive technique inducing transient peripheral hypoxia to subsequently enhance tissue tolerance against ischemia-reperfusion injury. This technique promotes local vasodilation, improves O~2~ delivery (Enko et al., [@B14]; Bailey et al., [@B1]), and enhances the efficiency of muscular contraction (Pang et al., [@B39]; Moses et al., [@B38]). Higher muscle O~2~ utilization was also demonstrated in the quadriceps of strength-trained male athletes after performing IPC compared to placebo compressions (Paradis-Deschenes et al., [@B40]). It is not surprising, therefore, to observe that IPC can improve maximal physical performance in various exercise modes in male participants (De Groot et al., [@B11]; Bailey et al., [@B1]; Paradis-Deschenes et al., [@B40]), although this is not a universal finding (Incognito et al., [@B23]). Before exercise sport performance applications, IPC was originally studied for its potential clinical relevance, and in this context, it is interesting to note that females appeared to display smaller clinical benefits compared with their male counterparts. In a cohort of 382 subjects, a failure to induce preconditioning effects during percutaneous coronary intervention was noted in females (Laskey and Beach, [@B31]), presumably due to the innate, multiple, protective actions of estrogens (Pitcher et al., [@B43]). Moreover, since physiological responses to exercise (in particular metabolic pathway utilization and perfusion) may differ between sexes, and that exercise performance data from females is almost inexistent, it is relevant and timely to examine the impact of IPC on females compared to males to more precisely evaluate the potency of this technique for exercise performance applications.
Females have been reported to exhibit less fatigue than males during intense exercise (Parker et al., [@B41]; Hunter, [@B21]), but not under conditions of ischemia (Russ and Kent-Braun, [@B44]), suggesting that muscle perfusion and oxygenation may be involved in the sex-related difference (Russ and Kent-Braun, [@B44]; Clark et al., [@B9]; Hunter et al., [@B22]). Indeed, females exhibit greater vasodilation in the limbs during single knee extensions (Parker et al., [@B41]), and may have a greater proportional area of type I fibers (Simoneau et al., [@B45]; Staron et al., [@B46]) and greater capacity for utilizing oxidative metabolism than males (Kent-Braun et al., [@B25]). Thus, females exhibit a greater reliance on oxidative metabolism compared with males, which could challenge the ergogenic impact of IPC.
Prior evidence examining sex-related differences in the response to IPC during exercise is limited. Gibson et al. ([@B17]) reported no performance difference between male and female team-sport athletes during five repeated 6-s sprints after performing 3 × 5-min occlusions at 220 mmHg to both legs. However, the IPC procedure used in that study failed to improve performance in either sex, making the sex comparison moot. Another study demonstrated the positive impact of IPC (2 × 3-min at 220-mmHg) on recovery from squat jump test and running sprint performance 24 h after an initial, fatiguing session in males, but the female cohort was too small to draw any firm conclusion (Beaven et al., [@B6]). Considering the scarcity of studies and that none have attempted to measure relevant physiological responses to provide sex-specific mechanistic insights, any conclusion on the usefulness of IPC for female athletes cannot robustly be drawn.
The aim of the current investigation was therefore to determine the impact of IPC on muscle force and haemodynamics (blood volume and O~2~ extraction) derived from near-infrared spectroscopy in males vs. females during repeated maximal efforts separated with incomplete recovery periods. We chose isolated contractions specifically to avoid confounding effects that inspiratory muscle fatigue can have on limb blood flow and O~2~ uptake (Kayser et al., [@B24]).
Materials and methods {#s2}
=====================
Participants
------------
Strength-trained (power and weight lifters, cross-fit and taekwondo athletes) males (*n* = 9, age 25 ± 2 year; height 1.78 ± 0.02 m; weight 86.5 ± 4.9 kg) and females (*n* = 8, age 22 ± 1 year; height 1.66 ± 0.02 m; weight 60.8 ± 2.7 kg) volunteered to take part in this study. All performed 3--5 weight training sessions per week. All participants were non-smokers, free of health problems, did not use any medication, and were asked to avoid vigorous exercise, alcohol and caffeine 24 h before the tests. All but one female (who was amenorrheic) were tested in their follicular phase. Participants provided written informed consent after being informed of experimental procedures, associated risks and potential benefits. The study was approved by the Ethics committee of Université Laval, and adhered to the principles established in the Declaration of Helsinki.
Experimental design
-------------------
Participants visited the laboratory for one familiarization and two experimental trials. Resting heart rate and blood pressure (inclusion criteria \<140/100 mmHg) were taken prior to every trial. During the first visit, height, weight and thigh circumference were measured. Thigh circumference (males: 61.3 ± 1.9 cm; females: 57.2 ± 1.9 cm) was measured by the same experimenter, 1 cm under the gluteal line. Participants then completed a familiarization session with the experimental set-up, comprising one 3-min compression at a pressure of 200 mmHg, and a standardized warm-up consisting of 5 min of cycling on a Monark ergometer (Ergomedic 828 E) at 100 W. The warm-up was continued with 3--5 right-leg extensions on an isokinetic dynamometer (Kin-Com 500 H, Chattecx Corp., Hixson, TN) at 20°/s, with effort perception progressing from 3 to 9 out of a scale of 10. After 2 min of rest while seated on the dynamometer, participants completed three complete sets of the exercise protocol described below.
Following familiarization, participants were randomized into IPC or SHAM groups in a single-blind, crossover design. In both conditions, participants were seated comfortably on a bed with both legs outstretched, and a non-elastic nylon blood pressure cuff (WelchAllyn, Skaneateles Falls, NY, USA, width: 21 cm) was positioned around the right upper thigh under the gluteal line. In IPC, the cuff was rapidly inflated to 200 mmHg for 5 min, and this was repeated three times with each compression episode separated by 5 min of reperfusion (cuff release) in the same position. A plateau in the NIRS-derived deoxy-hemoglobin concentration signal (see NIRS procedure below) was observed in every subject by 5 min, and taken as a sign of effective occlusion and ischemia. In SHAM, the cuff was inflated to 20 mmHg. To minimize any placebo effect, participants were told that the purpose of the study was to compare the impact of two different cuff pressures that could both alter performance.
The familiarization session and experimental trials were separated by a minimum of 3 days to eliminate the potential effects of the second window of protection caused by IPC (Bolli, [@B8]), and a maximum of 7 days. All trials were performed at the same time of day for every participant to avoid potentially confounding circadian rhythm effects. The laboratory temperature was controlled and constant (20.31 ± 0.02°C) throughout all trials.
Exercise protocol
-----------------
The exercise protocol started 18.5 ± 0.1 min after the end of the last cycle of compression. Participants were seated in an upright position on the isokinetic dynamometer, and a strap was secured tightly across the pelvis. The right leg was fixed to the dynamometer with a strap above the ankle external malleoli, and the axis of rotation was aligned to the lateral femoral condyle of the knee joint.
The protocol consisted of five sets of 5 maximum voluntary knee extensions (60° range of motion from 80 to 20°; 0° corresponding to knee fully extended) at 20°/s angular velocity (one extension lasting \~3.0 s). Participants were instructed to contract as hard as they could throughout the extension, and were strongly encouraged during all contractions. Contraction was stopped during flexion when the dynamometer arm automatically returned to 80° at angular velocity of 120°/s (lasting less than 0.5 s), and started immediately after the return of the arm. Subjects rested quietly and relaxed for 30 s between each set and after the last set. After the exercise, participants moved back to the bed to perform an arterial occlusion with the cuff at 200 mmHg (\~3--5 min) to obtain a physiological calibration of the NIRS signals. The cuff pressure was released after the deoxy-hemoglobin signal had reached a plateau (see Near-infrared spectroscopy procedure below). Participants were also asked which condition between IPC and SHAM they felt had the greatest impact on their performance, and their verbal answer recorded.
The force produced by participants was measured with a force transducer connected at the end of the level arm of the dynamometer, which was calibrated according to the manufacturer\'s recommendations before every trial (manufacturer typical error 0.5%). The intra- and inter-day coefficient of variation for force obtained by the main experimenter was 2.4%. Force signals were analyzed in Matlab® between a starting point defined when velocity was ≥18°/s, angle was ≥80° and force was ≥100 N, and an end point when velocity was ≥18°/s and angle was ≥20°. In every set, peak and average force were calculated. Total force was then calculated as the sum of the average force produced in all sets. Percent force decrement across all sets was calculated as follows: 100 − (\[total force output/ideal force output\] × 100), where total and ideal force outputs are the sum of average force values from all sets and the highest average force was multiplied by five, respectively.
Near-infrared spectroscopy (NIRS)
---------------------------------
NIRS is a versatile, non-invasive methodology providing semi-quantitative measures of tissue oxygenation, and is easily applied to study a variety of tissue regions in individuals. NIRS quantifies the changes in hemodynamics from changes in the absorption of near-infrared light by oxyhemoglobin (HbO~2~) and deoxyhemoglobin (HHb) (Mccully and Hamaoka, [@B35]). With this technique, oxygenation can be measured in a discrete region of a tissue in a working physiological setting, which enhances specificity and has distinct advantages as compared to more cumbersome methods. Muscle tissue oxygenation measured by NIRS reflects the balance of O~2~ delivery to working muscles and muscle O~2~ consumption in capillary beds (De Blasi et al., [@B10]; Ferrari et al., [@B16]). Assessment of de- and re-oxygenation kinetics during and after dynamic exercise has become increasingly popular in recent years as a means to non-invasively assess the aerobic function of skeletal muscle. In the current protocol, muscle blood volume and oxygenation were assessed using a portable spatially resolved, dual wavelength NIRS apparatus (PortaMon, Artinis Medical Systems BV, Netherlands). The NIRS device was installed on the distal part of the right vastus lateralis belly (approximately 15 cm above the proximal border of the patella). Skinfold thickness was measured at the site of the application of the NIRS (males: 8.7 ± 0.9 mm; females: 10.7 ± 2.0 mm) using a Harpenden skinfold caliper (Harpenden Ltd) during the familiarization session, and was less than half the distance between the emitter and the detector (i.e., 20 mm). This thickness is adequate to let near-infrared light through muscle tissue (Mccully and Hamaoka, [@B35]). The skin was cleaned with an alcohol swab, and the device was fixed using double-sided stick disks and tape. Black bandages were used to cover the device to eliminate potentially interfering background light. The position of the apparatus was marked with an indelible pen for repositioning during the subsequent visit. The pressure cuff was positioned above the NIRS device, which did not affect the placement of the device during occlusions.
A modified form of the Beer-Lambert law, using two continuous wavelengths (760 and 850 nm) and a differential optical path length factor of 4.95, was used to calculate micromolar changes in tissue oxy-hemoglobin (Δ\[HbO~2~\]), deoxy-hemoglobin (Δ\[HHb\]) and total hemoglobin (Δ\[THb\] = \[HbO~2~\] + \[HHb\]; used as an index of change in regional blood volume). NIRS data were acquired at 10 Hz. At rest, once the signal was stabilized, 1 min of baseline values were analyzed pre IPC and SHAM treatments. Then, NIRS signals were analyzed 2-min post treatment for a duration of 1 min to assess the impact of IPC on resting blood volume (Δ\[THb\]~rest~, μM). During exercise, NIRS analysis was limited to Δ\[HHb\] since this variable is less sensitive than \[HbO~2~\] to perfusion variations and abrupt blood volume changes during contraction and recovery (De Blasi et al., [@B10]; Ferrari et al., [@B16]). The \[HHb\] signal was averaged over the last second of every contraction and over every set to obtain peak (Δ\[HHb\]~peak~, % arterial occlusion) and mean (Δ\[HHb\]~avg~, % arterial occlusion) O~2~ extraction, respectively. These \[HHb\] data were then normalized to express the magnitude of changes from baseline, and expressed in percentage of the maximal amplitude calculated during an arterial occlusion performed at the end of exercise. Contraction metabolic efficiency was calculated as the average force/Δ\[HHb\]~avg~ ratio. Finally, during recovery periods between exercise sets, the muscle reoxygenation rate (ΔReoxy, μM.s^−1^) was calculated as the rate of change in \[HHb\] from the end of the exercise set to the end of the subsequent recovery period (i.e., the recovery of \[HHb\]; Billaut and Buchheit, [@B7]). During this period, the amplitude of change in \[THb\] (Δ\[THb\]~rec~) was also analyzed.
Statistical analysis
--------------------
All data are reported as means ± standard error (SE) or percentage changes from SHAM. The IPC-SHAM differences within the same group and between sexes were analyzed using Cohen\'s effect sizes (ES) ± 90% confidence limits (Batterham and Hopkins, [@B5]; Hopkins et al., [@B20]). Except for Δ\[THb\]~rest~ and Δ\[THb\]~rec~, all variables were log-transformed prior to analysis (Hopkins et al., [@B20]). Magnitudes of difference between conditions were determined with an effect size of 0.2 set to evaluate the smallest worthwhile change. Standardized effects were classified as small (\>0.2), moderate (\>0.5) or large (\>0.8). The effect was deemed "unclear" if chances of having better/greater or poorer/lower change in performance and physiological variables were both \>5% (Batterham and Hopkins, [@B5]; Hopkins et al., [@B20]).
Results {#s3}
=======
All 17 participants met all criteria, completed the entire protocol, and tolerated the IPC procedure without complications. None of the participants could tell what condition produced the greatest change in performance.
Performance
-----------
Force parameters for males and females are displayed in Table [1](#T1){ref-type="table"} and Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}. After the IPC manoeuver, total force clearly increased in males (13.0%, ES 0.64, 0.37;0.90), but the change was trivial in females (2.3%, ES 0.10, −0.17;0.38). &&&The sex difference for this parameter was clear (ES 0.40, 0.10;0.70; Figure [1](#F1){ref-type="fig"}). Specifically, the IPC-induced increase in average force was greater in males than females in every set of the protocol (set 1--males: 15.2% vs. females: 0.7%, ES 0.53, 0.16;0.90, set 2--males: 15.6% vs. females: 3.6%, ES 0.43, 0.10;0.76, set 3--males: 11.1% vs. females: 2.6%, ES 0.31, 0.02;0.61, set 4--males: 14.5% vs. females: 3.1%, ES 0.41, 0.07;0.76, set 5--males: 8.4% vs. females: 1.7%, ES 0.25, −0.06;0.56; Figure [2](#F2){ref-type="fig"}).
######
**Performance variables in IPC and SHAM conditions for males and females**.
**Females** **Males** **Sex difference**
--------------------- -------------- -------------- ---------------------------------- -------------- -------------- ------------------------------- -------------------------
Peak force S1 (N) 449.1 ± 37.2 481.3 ± 42.1 7.2%, ES 0.24^\*^, −0.07;0.56 627.9 ± 45.1 685.3 ± 35.7 10.2%, ES 0.41^\*^, 0.23;0.59 ES 0.10, −0.24;0.44
Peak force S2 (N) 421.1 ± 32.8 446.3 ± 33.1 6.4%, ES 0.22^\*^, 0.00;0.43 566.3 ± 31.3 648.9 ± 47.6 13.7%, ES 0.54^\*^, 0.32;0.75 ES 0.24^\*^, −0.03;0.50
Peak force S3 (N) 402.5 ± 22.7 429.5 ± 28.3 6.3%, ES 0.22^\*^, 0.05;0.38 558.8 ± 35.7 581.6 ± 41.7 3.8%, ES 0.16, −0.01;0.32 ES −0.09, −0.29;0.12
Peak force S4 (N) 402.5 ± 24.7 422.0 ± 25.0 5.1%, ES 0.18, −0.06;0.41 526.3 ± 29.8 574.6 ± 43.2 8.4%, ES 0.34^\*^, 0.06;0.61 ES 0.11, −0.20;0.42
Peak force S5 (N) 401.8 ± 27.1 397.5 ± 24.5 −0.7%, ES −0.02, −0.17;0.12 506.9 ± 28.3 544.0 ± 39.1 6.7%, ES 0.27^\*^, 0.01;0.54 ES 0.26^\*^, 0.00;0.51
Force decrement (%) 10.4 ± 1.5 8.4 ± 1.4 −19.8%, ES −0.38^\*^, −0.77;0.01 13.7 ± 2.7 15.2 ± 2.2 6.0%, ES0.12, −0.54;0.77 ES 0.45^\*^, −0.16;1.07
*Values are mean ± SE*.
*Asterisks (^\*^) denote "clear" effect sizes (see statistics section)*.
{#F1}
{#F2}
Importantly, although IPC did not increase average force in females, it clearly augmented 1-s peak force in sets 1--3 (Table [1](#T1){ref-type="table"}), whereas males displayed benefits in sets 1, 2, 4, and 5. This effect was higher in males compared to females in sets 2 and 5.
Percent force decrement was attenuated in females after IPC (−19.8%, ES −0.38, −0.77;0.01), which was clearly different than males (sex difference: ES 0.45, −0.16;1.07).
Muscle hemodynamics and oxygenation
-----------------------------------
Physiological variables are displayed in Table [2](#T2){ref-type="table"}. No difference was observed between IPC and SHAM in both groups for NIRS variables during baseline before the manoeuver. IPC increased Δ\[THb\]~rest~ to the same extent in males and females (males: 1.5% vs. females: 1.5%, ES 0.00, −0.80;0.79). After IPC, Δ\[THb\]~rec~ was increased in males after sets 1 (1.2%, ES 0.33, −0.14;0.80) and 5 (1.5% ES 0.44, 0.17;0.70). In females, Δ\[THb\]~rec~ was only enhanced after sets 2 (0.7%, ES 0.23, −0.09;0.54) and 3 (1.8% ES 0.54, 0.04;1.04). This change in Δ\[THb\]~rec~ was higher in males compared to females in set 5 only (sex difference: ES 0.48, 0.10;0.87). There was no change in ΔReoxy within the same group and between sexes.
######
**Muscle hemodynamic and oxygenation variables in IPC and SHAM conditions for males and females**.
**Females** **Males** **Sex difference**
------------------------- ------------- ------------- -------------------------------- ------------ ------------ -------------------------------- ------------------------
\[HHb\]~base~ (μM) 25.3 ± 1.3 25.2 ± 1.1 −0.20%, ES −0.01, −0.35;0.33 37.5 ± 2.7 37.8 ± 2.8 0.85%, ES 0.03, −0.14;0.21 −5.39;7.92
\[THb\]~base~ (μM) 46.1 ± 3.5 43.8 ± 3.3 −4.80%, ES −0.21, −0.50;0.08 72.2 ± 5.3 72.6 ± 5.6 0.42%, ES 0.02, −0.12;0.15 ES 0.16, −0.06;0.38
Δ\[THb\]~rest~ (μM) 0.1 ± 0.4 1.6 ± 0.5 1.5%, ES 1.06^\*^, 0.13;1.99 3.6 ± 0.6 5.1 ± 0.8 1.5%, ES 0.69^\*^, −0.17;1.55 ES 0.00, −0.80;0.79
Δ\[THb\]~rec~ S1 (μM) 7.3 ± 1.0 7.8 ± 0.7 0.53%, ES 0.16, −0.27;0.59 9.1 ± 1.1 10.3 ± 1.3 1.16%, ES 0.33^\*^, −0.14;0.80 ES 0.21, −0.46;0.88
Δ\[THb\]~rec~ S2 (μM) 5.0 ± 0.8 5.8 ± 0.5 0.75%, ES 0.23^\*^, −0.09;0.54 6.4 ± 0.9 6.5 ± 1.5 0.14%, ES 0.04, −0.50;0.58 ES −0.20, −0.88;0.48
Δ\[THb\]~rec~ S3 (μM) 3.7 ± 0.8 5.5 ± 0.8 1.77%, ES 0.54^\*^, 0.04;1.04 6.0 ± 1.0 6.5 ± 1.4 0.57%, ES 0.16, −0.21;0.53 ES −0.40, −1.04;0.25
Δ\[THb\]~rec~ S4 (μM) 4.7 ± 1.1 5.2 ± 0.7 0.58%, ES 0.18, −0.24;0.59 5.3 ± 1.1 5.9 ± 1.0 0.64%, ES 0.18, −0.26;0.62 ES 0.02, −0.62;0.66
Δ\[THb\]~rec~ S5 (μM) 5.5 ± 0.9 5.6 ± 0.77 0.08%, ES 0.02, −0.23;0.28 3.9 ± 0.9 5.4 ± 1.1 1.54%, ES 0.44^\*^, 0.17;0.70 ES 0.48^\*^, 0.10;0.87
ΔReoxy (μM.s^−1^) 0.1 ± 0.0 0.1 ± 0.02 −21.0%„ ES −0.32, −0.89;0.26 0.2 ± 0.1 0.2 ± 0.1 −4.0%„ ES −0.05, −0.38;0.028 ES 0.21, −0.29;0.70
Δ\[HHb\]~peak~ S1 (%AO) 59.0 ± 5.5 45.8 ± 13.3 −29.5%, ES −0.97, −2.86;0.92 69.8 ± 5.4 73.9 ± 3.9 7.3%, ES 0.26, −0.38;0.90 ES 1.35, −0.92;3.61
Δ\[HHb\]~peak~ S2 (%AO) 72.6 ± 7.3 57.3 ± 11.7 −10.3%, ES −0.30, −0.91;0.31 79.0 ± 5.6 83.9 ± 4.3 7.4%, ES 0.27, −0.35;0.89 ES 0.58, −0.24;1.40
Δ\[HHb\]~peak~ S3 (%AO) 75.0 ± 7.8 60.7 ± 11.8 −42.8%, ES −1.55, −4.11;1.01 80.4 ± 5.7 85.5 ± 3.9 7.7%, ES 0.28, −0.36;0.92 ES 2.04, −0.99;5.06
Δ\[HHb\]~peak~ S4 (%AO) 71.7 ± 7.3 61.4 ± 11.4 −29.1%, ES −0.95, −2.42;0.51 79.7 ± 6.8 86.3 ± 4.2 10.5%, ES 0.38, −0.24;0.99 ES 1.43, −0.32;3.17
Δ\[HHb\]~peak~ S5 (%AO) 68.4 ± 5.7 58.8 ± 11.5 −33.7%, ES −1.14, −2.64;0.36 82.7 ± 7.1 86.2 ± 4.2 6.4%, ES 0.23, −0.40;0.86 ES 1.52, −0.27;3.31
CMER S1 10.5 ± 1.4 5.6 ± 4.3 −4.4%, ES −0.12, −1.84;1.59 13.2 ± 2.5 11.8 ± 1.2 −2.4%, ES −0.04, −0.49;0.41 ES 0.05, −1.42;1.52
CMER S2 5.7 ± 1.3 7.2 ± 9.2 23.3%, ES 0.57, −0.57;1.72 5.6 ± 0.6 6.1 ± 0.8 8.7%, ES 0.14, −0.21;0.49 ES −0.29, −1.29;0.72
CMER S3 5.1 ± 1.1 9.9 ± 5.7 22.2%, ES 0.55, −0.46;1.56 5.1 ± 0.6 5.2 ± 0.6 3.8%, ES 0.06, −0.24;0.37 ES −0.37, −1.26;0.52
CMER S4 4.9 ± 1.0 17.3 ± 8.8 72.4%, ES 1.49, −0.42;3.41 4.7 ± 0.6 4.9 ± 0.5 5.0%, ES 0.08, −0.25;0.41 ES −1.13, −2.75;0.49
CMER S5 4.8 ± 0.9 19.3 ± 11.4 70.5%, ES 1.46, −0.30;3.23 4.6 ± 1.8 4.5 ± 0.5 −1.6%, ES −0.03, −0.36;0.30 ES −1.25, −2.75;0.25
*Values are mean ± SE*.
*Asterisks (^\*^) denote "clear" effect sizes (see statistics section). AO, arterial occlusion; CMER, contraction metabolic efficiency ratio; base, baseline before the IPC manoeuver; ΔReoxy, reoxygenation rate of the muscle; S, sets*.
The IPC maneuver did not alter muscle Δ\[HHb\]~peak~ across sets nor between sexes (Table [2](#T2){ref-type="table"}). However, Δ\[HHb\]~avg~ were higher after IPC in males for set 1 (18.1%, ES 0.31, −0.11;0.73), and lower in females for sets 3 (−44.4%, ES −0.25, −0.61;0.11), 4 (−40.2%, ES −0.22, −0.52;0.08), and 5 (−40.4%, ES −0.22, −0.50;0.06). There was a clear sex difference in the impact of IPC on global muscle deoxygenation in sets 2--5 (Figure [3](#F3){ref-type="fig"}). The contraction metabolic efficiency ratio was not altered by IPC during the sets or between sexes (Table [2](#T2){ref-type="table"}).
![**Average changes in (Δ\[HHb\]~**avg**~), expressed as a fraction of the maximal values obtained during a transient arterial occlusion (AO) in males (SHAM: ■ IPC: □) and females (SHAM: ● IPC: ○)**. *Asterisks (*^\*^*)* denote "clear" differences between sexes (see statistics section). Values are mean ± SE.](fphys-07-00674-g0003){#F3}
Discussion {#s4}
==========
Summary of main findings
------------------------
This study investigated the impact of sex on performance and vasoactive and oxidative responses to IPC in strength-trained athletes during repeated, maximal contractions. The main findings were that IPC (1) increased muscle force to a greater extent in males than females; (2) increased resting blood volume similarly in both sexes; and (3) increased O~2~ extraction in males but decreased it in females. These results challenge the general applicability of IPC on physical performance. While it may be recommended to enhance exercise capacity in males, this preconditioning technique appears less effective in females during maximal efforts.
Muscle force parameters
-----------------------
IPC affected muscle force production and the ability to resist neuromuscular fatigue differently between sexes. While male athletes increased peak and average maximal concentric force in every set of the protocol following IPC, the benefits in females were lesser (Figures [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}). Females only displayed small improvements in peak force in the first three sets, while average force changes were trivial across all sets. Importantly, while IPC yielded acute positive responses in every male, four females out of eight experienced a decrease in performance. Another study did not report any sex difference in repeated-sprint performance, but the IPC protocol employed did not yield any positive effects in either males or females and, therefore, sex-related differences could not be truly assessed (Gibson et al., [@B17]). Although not studying sex differences *per se*, Gibson and colleagues (Gibson et al., [@B18]) reported altered 30-m running sprint times after IPC in females, while males displayed no added benefit. Taken together, these results highlight the importance of considering inter-individual responses to IPC in sports, particularly in female athletes. Considering the complex sex modulation of preconditioning mechanisms (such as mitochondrial K~ATP~ channel activation, reactive oxygen species generation, nitric oxide synthase activity, and inflammatory mediator production), as well as robust data documenting that females experience an innate protective mechanism after several forms of acute injury (for review see Pitcher et al., [@B43]), one could expect a sex-specific impact of remote IPC on physical performance. That said, both sexes are capable of being preconditioned (Pitcher et al., [@B43]). Hence, it is possible that females in the current study did not reach a sufficient threshold for preconditioning to occur. Thus, contrary to the proposition of a responder vs. non-responder pattern (Beaven et al., [@B6]; Gibson et al., [@B18], [@B17]), the current data coupled with other clinical studies rather suggest that females may require a greater stimulus for effect. This remains to be elucidated by investigating the influence of varying numbers of IPC cycles and/or the number of limbs occluded at one time.
The sex-specific impact of IPC on fatigability has not been robustly assessed in the literature. The present data demonstrated clear differences between sexes; compared with males, females exhibited a lower force decrement over the five sets of maximal, isokinetic contractions. The lack of change in males is in keeping with previous studies reporting no differences between SHAM and IPC conditions in the measured fatigue index, despite higher peak and mean power outputs during the first repetitions of a series of ten 6 s cycle sprints (Patterson et al., [@B42]), or higher average force during maximal voluntary knee extensions after IPC (Paradis-Deschenes et al., [@B40]). This apparent sex difference in response to IPC could be attributed in part to the fact that males increased their initial and total force leading to greater subsequent metabolic and ionic perturbations (Balsom et al., [@B3]; Glaister, [@B19]). However, this cannot be the only explanation, as females clearly improved their resistance to fatigue across the sets by approximately 2%. Sex differences in the availability and use of O~2~ could shed some light on these differing responses to IPC.
Muscle hemodynamic and oxygenation
----------------------------------
IPC is known to increase blood flow in both ipsilateral (Kraemer et al., [@B30]) and contralateral limbs (Enko et al., [@B14]). It also up-regulates endothelial function at rest (Moro et al., [@B37]), after local transient ischemia (Kharbanda et al., [@B26]; Loukogeorgakis et al., [@B33]), and prevents the decline in flow-mediated dilation observed after strenuous exercise (Bailey et al., [@B1]) in males. By investigating the NIRS-derived \[THb\] changes from pre- to post-IPC, we confirmed the acute increase in local blood volume at rest in males, and extended this hyperperfusion finding to females. In fact, these moderate to large hemodynamic changes from baseline were similar in both sexes, of a magnitude of 1.5%. Such data from female participants are very scarce in the literature, and results do vary. While Kharbanda and colleagues reported no sex difference in flow-mediated dilation response during ischaemia-reperfusion after IPC (Kharbanda et al., [@B26]), the same response was found to be higher in females compared with males immediately post-preconditioning (Moro et al., [@B37]). There is, however, stronger evidence of a sex difference in vasodilator responsiveness. Females display higher brachial artery flow-mediated dilation (Levenson et al., [@B32]) and forearm vasodilatory response to acetylcholine and β2-adrenergic receptor stimulation (Dietz, [@B13]; Kneale et al., [@B29]) relative to males. At first glance, these sex-based differences in intrinsic physiological responses could explain the differing impact of IPC on performance in males vs. females observed in the current study. Greater effects on endothelial function should improve contractile activity (and/or efficiency) by allowing a better O~2~ supply to skeletal muscles during intense exercise. It is not clear why males had a greater increase in muscle force than females for a similar percent increase in Δ\[THb\]~rest~, but one could argue that males might benefit more than females from an up-regulated endothelial function since they possess a lower intrinsic vasodilator responsiveness. Nevertheless, NIRS does not offer a robust assessment of blood flow since it does not detect change in blood velocity (Delorey et al., [@B12]). Studies using Doppler ultrasound are warranted to investigate vasodilation and potential blood flow changes following IPC in males and females.
An augmented blood volume before exercise could facilitate O~2~ delivery to active skeletal muscles. While IPC-induced changes in peak muscle O~2~ extraction were trivial in males, they displayed meaningful changes in Δ\[HHb\]~avg~ across the sets (Figure [3](#F3){ref-type="fig"}). Males extracted more O~2~ than females after IPC in sets 2--5, with clearly increased muscle force. IPC has been reported to increase systemic maximal O~2~ uptake (De Groot et al., [@B11]), as well as local tissue deoxygenation at task failure during handgrip exercise at 45% maximal voluntary contraction (Barbosa et al., [@B4]), and decrease blood lactate accumulation during submaximal running exercise (Bailey et al., [@B2]). IPC also accelerates muscle deoxygenation dynamics and enhances performance during whole-body cycling and sustained isometric contraction of the knee in males (Kido et al., [@B27]; Tanaka et al., [@B47]). However, despite large Δ\[THb\]~rest~, IPC decreased Δ\[HHb\]~avg~ in females in the current study. The above studies exclusively recruited males, thus the current study adds to the literature by demonstrating that IPC might not induce similar metabolic responses in female athletes. Although we did not measure blood flow *per se*, the similar Δ\[THb\]~rest~ after IPC in both sexes suggest that the sex difference in O~2~ extraction might not be directly related to a difference in O~2~ availability. Furthermore, intramuscular pressure is positively correlated with contraction intensity, and it is accepted that occlusion of muscle blood flow occurs at 50--60% maximal voluntary contraction (Wigmore et al., [@B48]), thereby limiting its impact on sex differences observed in O~2~ metabolism when contractions are performed maximally as in the current study. In fact, the excessive intramuscular pressures of the contractions made our \[THb\] data collected during contractions unusable. And along this line of reasoning, it is of note that sex differences in fatigue development and performance disappear when blood flow is occluded (Maughan et al., [@B34]; Yoon et al., [@B51]). This could suggest that O~2~ availability does not explain the current finding of upregulated O~2~ extraction in males only. Caution is of course needed when interpreting NIRS-derived results due to methodological confounding factors such as subcutaneous fat layer thickness (although it was below the recommended emitter-receptor distance in both sexes) and possible NIRS sensor movement on the skin. Finally, muscle re-perfusion and thereby re-oxygenation occurring between maximal efforts is correlated with the recovery of muscle performance mainly via the resynthesis of phosphocreatine and by-products removal (Kime et al., [@B28]; Billaut and Buchheit, [@B7]). However, sex differences in Δ\[THb\]~rec~ were mostly unclear, as was the case for ΔReoxy. Based on these data, the sex-specific impact of IPC on exercise performance does not appear to be attributed to recovery processes.
There is also the possibility of a preferential impact of IPC on type II muscle fibers. A lower proportional area of type I fibers has been found in the vastus lateralis of males compared with females (Simoneau et al., [@B45]; Staron et al., [@B46]). Type II fibers display greater fractional O~2~ extraction with faster kinetics and lower microvascular O~2~ partial pressure (i.e., better muscle O~2~ diffusion index), despite a lower overall O~2~ consumption (Mcdonough et al., [@B36]). Interestingly, extraneous infusion of adenosine, which is a key acting molecule released during IPC, preferentially enhances vasodilation of arterioles to type II fibers (Wunsch et al., [@B50]). Therefore, due to the greater O~2~ extraction of type II fibers when highly perfused (Wilson et al., [@B49]), it has been suggested that these fibers might benefit more from an increase in blood perfusion than the more aerobic type I fibers (Faiss et al., [@B15]; Paradis-Deschenes et al., [@B40]), which could explain the higher Δ\[HHb\]~avg~ in strength-trained males in the present study.
In conclusion, this applied study demonstrated that strength-trained males might benefit more clearly from IPC than their female counterparts during repeated, maximal contractions. This strengthens clinical observations that sex may be a confounder in the response to this stimulus. Despite a similar increase in blood volume (↑\[THb\]) in both sexes immediately after IPC, and thus presumably similar increase in O~2~ availability, males displayed greater peripheral O~2~ extraction (↑Δ\[HHb\]). This ergogenic effect therefore appears to be mediated in part via an upregulation of oxidative function in males, possibly within type II muscle fibers.
Author contributions {#s5}
====================
PPD, DRJ, and FB conceptualized and designed the research project; PPD acquired the data and conducted the statistical analysis; PPD interpreted results with assistance from DRJ and FB; PPD wrote the manuscript with revisions from DRJ and FB. All authors reviewed and agreed upon 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.
The authors thank the athletes for their participation in the study. We also sincerely thank Pr Normand Teasdale and Mr. Marcel Kaszap for their valuable technical support and assistance.
[^1]: Edited by: Gregoire P Millet, University of Lausanne, Switzerland
[^2]: Reviewed by: Samuel Verges, Université Joseph Fourier, France; Philip J Millar, University of Guelph, Canada
[^3]: This article was submitted to Exercise Physiology, a section of the journal Frontiers in Physiology
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Speed-accuracy trade-offs (SATs), such as that between movement speed and spatial accuracy, are one of the most common phenomena in human movement. SATs have been actively studied since Woodworth^[@CR1]^, who measured the accuracy of voluntary movement. Fitts' law, the most famous SAT model, has been widely accepted in the field of Ergonomics and Human-Computer Interaction. Fitts^[@CR2]^ applied Shannon's information theory to the human motor system and proposed an empirical model that relates movement duration to movement distance and target width, called Fitts' law, based on the following equation:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$D=a+b\,{\log }_{2}(\frac{2A}{W})=a+b\,{I}_{d},$$\end{document}$$where *D* and *A* are the movement duration and distance, respectively, *W* is the target width, *I*~*d*~ is a logarithmic term in *W* and *A* called the index of difficulty, and *a* and *b* are intercept and slope in the linear equation, respectively, which are obtained as regression coefficients using *D* as an objective variable and *I*~*d*~ as an explanatory variable. Equation ([1](#Equ1){ref-type=""}) can be rewritten as follows:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$D=a^{\prime} +b\,{\log }_{2}(\frac{1}{W}),\,a^{\prime} =a+b({\log }_{2}A+1),$$\end{document}$$to emphasise that increasing the spatial accuracy required (i.e., decreasing *W*) at a given distance increases movement duration, and vice versa. Thus, Fitts' law expresses a linear SAT function represented by features *a* and *b*, i.e., intercept and slope. Previous studies have already shown that Fitts' law can be applied to behavioural data in a variety of contexts (reviewed in^[@CR3]--[@CR5]^). There is also a considerable body of research supporting the applicability of Fitts' law to arm movements such as reciprocal tapping^[@CR2]^, reaching movements^[@CR6]^, and one-dimensional linear and two-dimensional planar arm movements^[@CR7]--[@CR10]^.
As mentioned above, Fitts' law is a model that has been supported by many researchers for more than half a century, but the question as to how Fitts' law can explain human movement under any condition, despite being an empirical model, remains unanswered. From a physical perspective, human movements are modelled by the fundamental law of motion, that is, using dynamics (including kinematics). Therefore, we hypothesised that Fitts' law in reaching movement could be determined in terms of human arm dynamics. It is important to clarify the relationship between speed-accuracy and arm dynamics because it enables us to understand human motor performance at a more profound level.
There have been interesting findings regarding the slope of Fitts' law, including demonstrations that the slope increases as the size of effectors^[@CR11]^, the slope in elderly subjects is greater than that in young adults^[@CR12]--[@CR14]^, and the slope can be reduced with practice^[@CR15]^. From these experimental facts, it is quite natural to consider that the coefficients of Fitts' law are related to the physical dynamics factors. Recent studies have further shown that it is essential to investigate dynamics factors. For example, Hoffmann and Hui^[@CR16]^ investigated the duration taken to move a given distance using different arm components such as fingers, wrists, forearms and the full arm and concluded that the moments of inertia and muscle torque strength of these components affect movement duration. Bertucco *et al*.^[@CR17]^ showed that movements over a long distance were associated with Coriolis forces and had a larger intercept of Fitts' law than those at short distances. Although these studies showed that dynamics factors such as arm segmentation and torque are involved in the SAT profiles of Fitts' law, no sufficient theoretical explanation was provided. According to dynamical optimisation models such as the minimum (commanded) torque change criterion^[@CR18],[@CR19]^, increasing movement duration changes the proportions of the respective torque components (inertial, Coriolis, centrifugal and viscous forces) along with the movement trajectory. To better understand the SAT mechanism in human arm movement, it is therefore important to construct a model that considers the contribution of these arm dynamics factors to movement.
Here, we derive the relationship between movement duration *D* and spatial endpoint error *W*---the factors of Fitts' law---based on multi-joint forward- and inverse-dynamics models. As some researchers have pointed out that modelling the SAT requires consideration of the signal-dependent biological noise in the nervous system^[@CR20]--[@CR22]^, we consider modelling in the presence of signal-dependent noise. The *D* − *W* relationship that we derive nearly covers Fitts' law. Furthermore, we theoretically demonstrate that arm dynamics and noise factors can affect both the slope and intercept of Fitts' law.
Explanation of Fitts' Law Based on Human Arm Dynamics {#Sec2}
=====================================================
To formulate the relationship between movement duration *D* and hand endpoint error *W* based on human arm dynamics, we assumed the feedforward control model shown in Fig. [1](#Fig1){ref-type="fig"}, and modelled the relationship between *D* and *W* using the following three steps: first, we derived the relationship between movement duration *D* and joint torque *τ* based on an inverse dynamics model; next, we derived the relationship between *τ* and torque noise *τ*^noise^ by defining a signal-dependent noise; finally, we derived the relationship between *τ*^noise^ and hand endpoint error *W* using a forward dynamics model. We then used the resulting *D* − *W* relation to represent Fitts' law.Figure 1Schematic of modelling, involving feedforward control of rapid arm-reaching movement with movement planning and execution performed sequentially. The movement duration *D* is determined in the task selection stage. The desired trajectory is planned in the planning stage, in which the desired torque *τ*(*t*) at time *t*, which corresponds to the command signal, is generated by the feedforward controller implementing an inverse dynamics model. Finally, the torque noise *τ*^noise^(*t*) is added at the movement execution stage. The hand trajectory is generated by the controlled object (i.e., the arm), which implements a forward dynamics model using the noise-added torque *τ*(*t*) + *τ*^noise^(*t*). The result of converting the torque noise into the task space is the spatial error of the movement, and its endpoint *t* = *D* is the hand endpoint error.
Step 1: Determining the relationship between movement duration and joint torque {#Sec3}
-------------------------------------------------------------------------------
The general form of multi-joint (*N*-link) human arm dynamics on the horizontal plane can be modelled as a second-order nonlinear differential equation^[@CR18]^:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{\tau }}(t)={\boldsymbol{M}}({\boldsymbol{\theta }}(t))\mathop{{\boldsymbol{\theta }}}\limits^{{\boldsymbol{..}}}(t)+{{\boldsymbol{h}}}_{1}({\boldsymbol{\theta }}(t))[\dot{{\boldsymbol{\theta }}}(t)\dot{{\boldsymbol{\theta }}}(t)]+{{\boldsymbol{h}}}_{2}({\boldsymbol{\theta }}(t))[\dot{{\boldsymbol{\theta }}}{(t)}^{2}]+{\boldsymbol{B}}\dot{{\boldsymbol{\theta }}}(t),$$\end{document}$$where *t* represents time, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{\tau }}(t)={[{\tau }_{1}(t),{\tau }_{2}(t),\cdots ,{\tau }_{i}(t),\cdots ,{\tau }_{N}(t)]}^{{\rm{T}}}$$\end{document}$ is the joint torque, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{\theta }}(t)=[{\theta }_{1}(t),{\theta }_{2}(t),\cdots ,$$\end{document}$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\theta }_{i}(t),\cdots ,{\theta }_{N}(t)]}^{{\rm{T}}}$$\end{document}$ is the joint angular position, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\dot{{\boldsymbol{\theta }}}(t)$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\dot{{\boldsymbol{\theta }}}(t)$$\end{document}$ are the joint angular velocity and acceleration, respectively, the superscript T denotes the transpose of a vector or matrix, and the subscript *i* denotes the joint number. The first term is the inertial force given in terms of the inertia matrix $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{M}}({\boldsymbol{\theta }}(t))\in {{\mathbb{R}}}^{N\times N}$$\end{document}$. The second term is the Coriolis force, which is given in terms of a Coriolis force coefficient matrix $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\boldsymbol{h}}}_{1}({\boldsymbol{\theta }}(t))\in {{\mathbb{R}}}^{N\times N(N-1)/2}$$\end{document}$. The third and fourth terms are the centrifugal and viscous forces in terms of the coefficient matrices of centrifugal and viscous forces, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\boldsymbol{h}}}_{2}({\boldsymbol{\theta }}(t))\in {{\mathbb{R}}}^{N\times N}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{B}}\in {{\mathbb{R}}}^{N\times N}$$\end{document}$, respectively. Further, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$[\dot{{\boldsymbol{\theta }}}(t)\dot{{\boldsymbol{\theta }}}(t)]$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$[\dot{{\boldsymbol{\theta }}}{(t)}^{2}]$$\end{document}$ are the vectors in Coriolis and centrifugal forces, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${[{\dot{\theta }}_{1}(t){\dot{\theta }}_{2}(t),{\dot{\theta }}_{1}(t){\dot{\theta }}_{3}(t),\cdots ,{\dot{\theta }}_{N-1}(t){\dot{\theta }}_{N}(t)]}^{{\rm{T}}}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${[{\dot{\theta }}_{1}{(t)}^{2},{\dot{\theta }}_{2}{(t)}^{2},\cdots ,{\dot{\theta }}_{N}{(t)}^{2}]}^{{\rm{T}}}$$\end{document}$, respectively.
As the hand path observed in human arm movement is nearly invariant within a specific range of movement duration^[@CR23]^, the following time-normalisation can be assumed^[@CR24]^ (for a validation of the time-normalisation, see Discussion and Supplementary Fig. [S1](#MOESM1){ref-type="media"}):$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$t=Ds,\,0\le t\le D\,(0\le s\le 1),\,{\theta }_{i}(t)={\theta }_{i}(Ds)={\tilde{\theta }}_{i}(s),$$\end{document}$$where *s* is the normalised time. Equation ([3](#Equ3){ref-type=""}) can be further expressed in terms of the movement duration *D* as$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\begin{array}{rcl}{\boldsymbol{\tau }}(t) & = & {\boldsymbol{\tau }}(Ds)=\frac{1}{{D}^{2}}\{{\boldsymbol{M}}(\tilde{{\boldsymbol{\theta }}}(s))\mathop{\tilde{{\boldsymbol{\theta }}}}\limits^{{\boldsymbol{..}}}(s)+{{\boldsymbol{h}}}_{1}(\tilde{{\boldsymbol{\theta }}}(s))[\dot{\tilde{{\boldsymbol{\theta }}}}(s)\dot{\tilde{{\boldsymbol{\theta }}}}(s)]\\ & & +\,{{\boldsymbol{h}}}_{2}(\tilde{{\boldsymbol{\theta }}}(s))[\dot{\tilde{{\boldsymbol{\theta }}}}{(s)}^{2}]\}+\frac{1}{D}{\boldsymbol{B}}\dot{\tilde{{\boldsymbol{\theta }}}}(s).\end{array}$$\end{document}$$
We can further assume that the approximation $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$1/D\simeq \lambda /{D}^{2}$$\end{document}$ is possible over the general movement duration range, where *λ* is the coefficient of approximation (see Supplementary Fig. [S2](#MOESM1){ref-type="media"}), giving:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\begin{array}{c}{\boldsymbol{\tau }}(Ds)\simeq \frac{1}{{D}^{2}}\{{\boldsymbol{M}}(\tilde{{\boldsymbol{\theta }}}(s))\mathop{\tilde{{\boldsymbol{\theta }}}}\limits^{{\boldsymbol{..}}}(s)+{{\boldsymbol{h}}}_{1}(\tilde{{\boldsymbol{\theta }}}(s))[\dot{\tilde{{\boldsymbol{\theta }}}}(s)\dot{\tilde{{\boldsymbol{\theta }}}}(s)]+{{\boldsymbol{h}}}_{2}(\tilde{{\boldsymbol{\theta }}}(s))[\dot{\tilde{{\boldsymbol{\theta }}}}{(s)}^{2}]+\lambda {\boldsymbol{B}}\dot{\tilde{{\boldsymbol{\theta }}}}(s)\},\\ \,\,\simeq \frac{1}{{D}^{2}}\tilde{{\boldsymbol{\tau }}}(s),\end{array}$$\end{document}$$where $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\tilde{{\boldsymbol{\tau }}}(s)$$\end{document}$ is a time-normalised torque not related to the movement duration. Expressions have been developed using the general notation (*N*-joint); the two-joint case for shoulder and elbow is described in detail in Supplementary Equation [1](#MOESM1){ref-type="media"} and Supplementary Fig. [S3](#MOESM1){ref-type="media"}.
Step 2: Determining the relationship between desired torque and torque noise {#Sec4}
----------------------------------------------------------------------------
The actual neural process of human movement is affected by biological noise. Therefore, in the movement execution stage, we assume the presence of a signal-dependent noise represented by Gaussian noise with a mean of zero and a standard deviation proportional to the control signal that additively acts on the control signal^[@CR20]^. Here, we refer to the signal-dependent noise at the torque level as the torque noise, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\tau }_{i}^{{\rm{noise}}}$$\end{document}$, which is represented as follows:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\tau }_{i}^{{\rm{noise}}}(t)={k}_{i}|{\tau }_{i}(t)|{z}_{i}(t),$$\end{document}$$where the noise parameter *k*~*i*~ indicates the degree to which the noise depends on the desired torque and *z*~*i*~(*t*) is a pseudo-random variable describing the standard normal distribution. Time-normalisation of Eq. ([7](#Equ7){ref-type=""}) gives$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\tau }_{i}^{{\rm{noise}}}(Ds)=\frac{1}{{D}^{2}}{k}_{i}|{\tilde{\tau }}_{i}(s)|{z}_{i}(s).$$\end{document}$$
Step 3: Determining the relationship between torque noise and hand endpoint error {#Sec5}
---------------------------------------------------------------------------------
The forward model is used to express a motor process in which endpoint error occurs as a result of the execution of a movement with noise-added torque. For instance, the forward model for a case of *x*-coordinate movement in a horizontal plane, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\boldsymbol{f}}}^{{\rm{FM}}}\in {{\mathbb{R}}}^{2\times 1}$$\end{document}$, can be used to obtain the state variable of the hand, including position and velocity, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{X}}(t)={[x(t),\dot{x}(t)]}^{{\rm{T}}}\in {{\mathbb{R}}}^{2\times 1}$$\end{document}$ from the joint torque $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\boldsymbol{\tau }}(t)={[{\tau }_{1}(t),{\tau }_{2}(t),\cdots ,{\tau }_{N}]}^{{\rm{T}}}\in {{\mathbb{R}}}^{N\times 1}$$\end{document}$ as follows:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\frac{{\rm{d}}}{{\rm{d}}t}{\boldsymbol{X}}(t)={{\boldsymbol{f}}}^{{\rm{FM}}}({\boldsymbol{X}}(t),{\boldsymbol{\tau }}(t)).$$\end{document}$$
Applying time integration from the beginning to the end of the movement (0 ≤ *t* ≤ *D*) to Eq. ([9](#Equ9){ref-type=""}) and taking the difference between the equations with and without noise yields (see Supplementary Equation [2](#MOESM1){ref-type="media"} for details):$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta {\boldsymbol{X}}(D)\simeq {\int }_{0}^{D}\frac{\partial {{\boldsymbol{f}}}^{{\rm{FM}}}({\boldsymbol{X}}(t),{\boldsymbol{\tau }}(t))}{\partial {\boldsymbol{\tau }}{(t)}^{{\rm{T}}}}\cdot {{\boldsymbol{\tau }}}^{{\rm{noise}}}(t){\rm{d}}t.$$\end{document}$$
The partial derivative of the forward model ***f*** ^FM^ with respect to the torque ***τ*** is then approximated as follows (see Supplementary Equation [3](#MOESM1){ref-type="media"} for details):$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\frac{\partial {{\boldsymbol{f}}}^{{\rm{FM}}}({\boldsymbol{X}}(t),{\boldsymbol{\tau }}(t))}{\partial {\boldsymbol{\tau }}{(t)}^{{\rm{T}}}}\simeq [\begin{array}{c}{{\boldsymbol{J}}}^{x}({\boldsymbol{\theta }}(t)){{\boldsymbol{B}}}^{-1}\\ {{\boldsymbol{J}}}^{x}({\boldsymbol{\theta }}(t)){\boldsymbol{M}}{({\boldsymbol{\theta }}(t))}^{-1}\end{array}],$$\end{document}$$where $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${{\boldsymbol{J}}}^{x}({\boldsymbol{\theta }}(t))$$\end{document}$ is a Jacobian matrix from the joint space to the *x*-coordinate space. From Eq. ([10](#Equ10){ref-type=""}), the positional endpoint error of hand state variable can be expressed in terms of the torque noise as follows:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta x(D)\simeq \mathop{\sum }\limits_{i=1}^{N}\,{\int }_{0}^{D}{({{\boldsymbol{J}}}^{x}({\boldsymbol{\theta }}(t)){{\boldsymbol{B}}}^{-1})}_{i}{\tau }_{i}^{{\rm{noise}}}(t){\rm{d}}t,$$\end{document}$$where $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${(\cdot )}_{i}$$\end{document}$ is the vector in the *i*-th column (*i*-th joint) in the matrix shown in parentheses. By time-normalising Eq. ([12](#Equ12){ref-type=""}) and substituting Eq. ([8](#Equ8){ref-type=""}), the following relation is obtained:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta x(D)\simeq \frac{1}{D}\mathop{\sum }\limits_{i=1}^{N}\,{k}_{i}{\alpha }_{i}^{x},\,{\alpha }_{i}^{x}={\int }_{0}^{1}{({{\boldsymbol{J}}}^{x}(\tilde{{\boldsymbol{\theta }}}(s)){{\boldsymbol{B}}}^{-1})}_{i}|{\tilde{\tau }}_{i}(s)|{z}_{i}(s){\rm{d}}s.$$\end{document}$$
Similar derivations to those in Eqs. ([9](#Equ9){ref-type=""}--[13](#Equ13){ref-type=""}) can be applied to the *y*-coordinate:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta y(D)\simeq \frac{1}{D}\mathop{\sum }\limits_{i=1}^{N}\,{k}_{i}{\alpha }_{i}^{y},\,{\alpha }_{i}^{y}={\int }_{0}^{1}{({{\boldsymbol{J}}}^{y}(\tilde{{\boldsymbol{\theta }}}(s)){{\boldsymbol{B}}}^{-1})}_{i}|{\tilde{\tau }}_{i}(s)|{z}_{i}(s){\rm{d}}s.$$\end{document}$$where $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{x}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{y}$$\end{document}$ are constants in terms of *D* and are the values calculated by the dynamics parameters of the arm.
The model endpoint error *W* ^model^ in the horizontal plane is then expressed as the Euclidean distance between Δ*x*(*D*) and Δ*y*(*D*):$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\begin{array}{rcl}{W}^{{\rm{model}}} & \equiv & \sqrt{\Delta x{(D)}^{2}+\Delta y{(D)}^{2}},\\ & = & \frac{\gamma }{D},\,\gamma =\sqrt{{(\mathop{\sum }\limits_{i=1}^{N}{k}_{i}{\alpha }_{i}^{x})}^{2}+{(\mathop{\sum }\limits_{i=1}^{N}{k}_{i}{\alpha }_{i}^{y})}^{2}},\end{array}$$\end{document}$$where *γ* is calculated using the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{x}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{y}$$\end{document}$, and *k*~*i*~. *k*~*i*~ is a signal-dependent noise parameter for each joint. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{x}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{y}$$\end{document}$ are calculated using the following dynamics parameters:Kinematic parameters: the time-normalised joint angular position $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\tilde{\theta }}_{i}(s)$$\end{document}$, velocity $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\dot{\tilde{\theta }}}_{i}(s)$$\end{document}$, and acceleration $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\ddot{\tilde{\theta }}}_{i}(s)$$\end{document}$ for each joint.Physical parameters: the *i*-th link length *L*~*i*~, distance from the joint to the centre of mass *S*~*i*~, mass *m*~*i*~, moment of inertia *I*~*i*~, and joint viscosity coefficient *B*~*ij*~.
Physical meanings of Fitts' law coefficients {#Sec6}
--------------------------------------------
As a further consideration, we consider the physical meanings of the Fitts' law coefficients, *a* and *b*. As these coefficients represent the relationship between speed and accuracy, they have been used as an index of human motor performance. Therefore, examining the coefficients of Fitts' law is an important step to gaining a deeper understanding of human motor performance.
Solving Eq. ([15](#Equ15){ref-type=""}) for *D* gives$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$D=\frac{\gamma }{{W}^{{\rm{model}}}}.$$\end{document}$$
To bring this equation into the Fitts' law form, we first consider a first-order Taylor expansion of ln (1 + *x*) around *x* = 0; $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\mathrm{ln}\,(1+x)\simeq x$$\end{document}$, for −1 \< *x* ≤ 1. By setting $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$x=\frac{c}{{W}^{{\rm{model}}}}-1$$\end{document}$, we obtain$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\mathrm{ln}(\frac{c}{{W}^{{\rm{model}}}})\simeq \frac{c}{{W}^{{\rm{model}}}}-1,\,{\rm{for}}\,0 < \frac{c}{{W}^{{\rm{model}}}}\le 2.$$\end{document}$$where *c* is a constant. From the change-of-base formula, the following linear approximation is obtained:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\begin{array}{ll}\frac{1}{{W}^{{\rm{model}}}}\simeq {p}_{1}+{p}_{2}\,{\log }_{2}(\frac{1}{{W}^{{\rm{model}}}}), & {\rm{for}}\,\frac{c}{2}\le {W}^{{\rm{model}}},\\ {p}_{1}=\frac{1}{c}(1+\,\mathrm{ln}\,c), & \\ {p}_{2}=\frac{1}{c\,{\log }_{2}e}, & {\rm{where}}\,e\,{\rm{is}}\,{\rm{Napier}}\mbox{'}{\rm{s}}\,{\rm{constant}}\,(e\simeq 2.71828\cdots ).\end{array}$$\end{document}$$
If *c* = 0.0060, the range of *W* ^model^ is 0.003 m ≤ *W* ^model^ and *p*~1~ and *p*~2~ are determined as −686.0 and 115.5, respectively. Once the value of *c* is determined, *p*~1~ and *p*~2~ are irrelevant to the subject or task. Finally, the approximation of the model to Fitts' law is represented by$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$D={a}^{{\rm{model}}}+{b}^{{\rm{model}}}\,{\log }_{2}(\frac{1}{{W}^{{\rm{model}}}}),$$\end{document}$$$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${a}^{{\rm{model}}}={p}_{1}\gamma ,$$\end{document}$$$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${b}^{{\rm{model}}}={p}_{2}\gamma .$$\end{document}$$
The above equations show that *γ* is equally involved in both *a*^model^ and *b*^model^, which correspond to the intercept and slope of Fitts' law, respectively. Therefore, the slope and intercept of Fitts' law can be expressed by arm dynamics and signal-dependent noise parameters. Previous studies have observed that the slope and intercept of Fitts' law are related to, for example, arm components (such as fingers, wrists, forearms and full arm) and arm physical parameters (such as length, mass, the moment of inertia)^[@CR11]--[@CR17]^ and, therefore, such empirical findings might be revealed theoretically by further study.
Because *c* is always positive, *p*~2~ is always positive; thus, increasing *γ* increases the Fitts' law slope, and vice versa. We therefore consider the case in which *γ* increases. From Eq. ([15](#Equ15){ref-type=""}), *γ* increases when $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$|{\sum }_{i=1}^{N}\,{k}_{i}{\alpha }_{i}^{x}|$$\end{document}$ or $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$|{\sum }_{i=1}^{N}\,{k}_{i}{\alpha }_{i}^{y}|$$\end{document}$ increases. Both terms contain the signal-dependent noise parameter *k*~*i*~, which is always positive, so that *γ* increases as *k*~*i*~ increases; that is, the slope of Fitts' law increases as the noise level increases. Focusing on the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{x}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{y}$$\end{document}$ calculated by the arm dynamics parameters, *γ* can also be reduced by reducing $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{x}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{y}$$\end{document}$, even if *k*~*i*~ has a large value. There are two ways to reduce $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{x}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\alpha }_{i}^{y}$$\end{document}$---reduce the absolute value of the time-normalised torque or reduce the product of the Jacobian matrix and the inverse viscosity matrix---but it is difficult to intuitively understand how to reduce these owing to the nonlinearity of the dynamics. To understand these factors, it is preferable to rely on numerical calculations such as sensitivity analysis.
Results {#Sec7}
=======
The relationship between movement duration and hand endpoint error could be modelled using *W* ^model^ = *γ*/*D*. We first investigated how the degree of hand endpoint error calculated by the model could represent the hand endpoint error observed in actual movement. We then examined how close the model function is to Fitts' law when it was applied to the Fitts' law axis, log~2~ (1/*W*) − *D*.
The behavioural experiment involved reaching the target at four points on the horizontal plane (front, back, left and right, as viewed from the starting point) within a given movement duration range. Eleven subjects participated, and the distance between the starting point and target point was fixed at 15 cm. The shoulder, elbow and hand positions were measured using a three-dimensional optical position measurement digitizer. The experiment was continued until a total of 320 successful trials (i.e., 80 trials in each direction) had been conducted (see Methods for details of the experiment).
How well the degree of model endpoint error can represent the actual endpoint error {#Sec8}
-----------------------------------------------------------------------------------
We examined how the degree of endpoint error calculated by the model can explain the endpoint error observed in actual movement. Figure [2](#Fig2){ref-type="fig"} shows a comparison between the observed and model endpoint errors in a representative subject (see Supplementary Figs. [S4](#MOESM1){ref-type="media"} to [S13](#MOESM1){ref-type="media"} for other subjects). Table [1](#Tab1){ref-type="table"} lists the root mean square errors (RMSEs) between the model and observed endpoint error for each subject. The model represented the trial-averaged observed endpoint error well, with a small average error of 0.0043 m. Subjects A, E and K had small RMSEs in all directions (average RMSE less than 0.0030 m), while C, F and H had high RMSEs in all directions (average RMSE greater than 0.0060 m). Overall, the model endpoint error closely matched the observed endpoint errors of actual movement. The observed and model endpoint errors for single-trial are not in one-to-one correspondence (the observed endpoint errors were measured for 20 trials per movement duration condition, while the model endpoint errors were generated over 100 trials using different pseudorandom numbers). The distributions of endpoint errors in the observed and model results were roughly consistent.Figure 2Comparison between observed and model endpoint errors for a representative subject (A). Tiny grey circles show observed endpoint error for each observed movement duration. Open circles show averaged values of observed endpoint errors at the mean observed movement duration. Tiny light blue crosses show the model endpoint errors at the mean observed movement duration using different pseudorandom numbers. When calculating this, the average of 20 measured trajectories in each movement duration was used as the time-normalised trajectory. Blue triangles show the trial-averaged values of the model endpoint error at the mean observed movement duration. The blue line shows *D* − *W* function of the model, which is obtained by using the mean value of *γ* over the entire movement duration (i.e., 0.30--0.70 s). In calculating this, the average of 80 measured trajectories over the entire movement duration was used as the time-normalised trajectory. Histograms hatched in translucent grey and blue are the distributions after fitting the observed and model endpoint error data with a Rayleigh distribution. As the *x*- and *y*-axis endpoint errors follow a Gaussian distribution, their Euclidean distance (i.e., *W*) follows a Rayleigh distribution.Table 1RMSE between model endpoint error and observed endpoint error for each subject.RMSE (m)SubjectABCDEFGHIJKMean ± SD\*Front0.00220.00280.00430.00350.00280.00870.00310.00400.00220.00430.00380.0038 ± 0.0017Back0.00190.00320.00500.00310.00180.01300.00280.00510.00310.00190.00280.0040 ± 0.0031Left0.00130.00370.00760.00680.00290.00730.00430.00710.00550.00320.00160.0047 ± 0.0022Right0.00120.00260.00890.00570.00430.00740.00330.00750.00670.00370.00270.0049 ± 0.0023Mean0.00170.00310.00640.00480.00290.00910.00340.00600.00440.00330.00270.0043 ± 0.0020\*SD: standard deviation across subjects.
Among the four directions, leftward endpoint variability was large while backward endpoint variability was small; overall, the model captured these direction-dependent features of hand endpoint variability. The direction dependency of the differences could be attributed to variations in the viscosity coefficient *B*~*ij*~ and time-normalised trajectories $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\tilde{\theta }(s)$$\end{document}$, velocities $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\dot{\tilde{\theta }}(s)$$\end{document}$, and accelerations $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\ddot{\tilde{\theta }}(s)$$\end{document}$, as all other physical parameters (*L*~*i*~, *S*~*i*~, *m*~*i*~, *I*~*i*~) and signal-dependent noise parameters (*k*~*i*~) were the same in all directions.
How close the model function is to Fitts' law {#Sec9}
---------------------------------------------
We investigated how close the model, *D* = *γ*/*W* ^model^, is to Fitts' law, *D* = *a*′ + *b*log~2~ (1/*W*) along the log~2~ (1/*W*) − *D* axis. Figure [3](#Fig3){ref-type="fig"} shows a simple conversion of the *D* − *W* axes shown in Fig. [2](#Fig2){ref-type="fig"} to log~2~ (1/*W*) − *D* axes. Fitts' law (black-dashed line) is indicated by a straight line, while the model line (blue solid line) is somewhat concave along the axis of Fitts' law, log~2~ (1/*W*) − *D*. This concavity has been observed in many previous studies^[@CR3],[@CR4],[@CR11],[@CR14],[@CR25]^. The functions are quite close to each other. Similar results were obtained for several other subjects (see Supplementary Figs. [S4](#MOESM1){ref-type="media"} to [S13](#MOESM1){ref-type="media"} for other subjects' results).Figure 3Comparison between Fitts' law and the model in *I*~*d*~ − *D* axis for a representative subject (A). The tiny grey and open circles, light blue crosses and blue triangles represent the same factors shown in Fig. [2](#Fig2){ref-type="fig"}. The black dashed and blue solid lines indicate the Fitts' law results (regression line for the trial-averaged *W* ^observed^) and the model line along the *I*~*d*~ − *D* axis, respectively.
A comparison of the parameters describing Fitts' law (*a*′ and *b*) with the corresponding parameters for the model (*a*^model^ and *b*^model^) revealed average slope values over the subjects and directions of *b* = 0.340 and *b*^model^ = 0.431, respectively, and average intercept values over the subjects and directions of *a*′ = −1.752 and *a*^model^ = −2.558, respectively. Thus, in terms of subject-averaged values, both *a*^model^ and *b*^model^ were close to the Fitts' law values of *a*′ and *b*, respectively. These results indicate that Fitts' law can be explained by the *W* − *D* relation derived based on human arm dynamics.
Discussion {#Sec10}
==========
The relationship between movement duration *D* and hand endpoint error *W*---the factors used in Fitts' law---were modelled based on human arm dynamics and found to follow an inversely proportional form: *W* = *γ*/*D* (where *γ* is a coefficient based on arm dynamics and signal-dependent noise parameters). The model was found to nearly coincided with Fitts' law when plotted on the axes log~2~ (1/*W*) and *D*. Furthermore, it was theoretically shown that the slope and intercept of Fitts' law include both arm dynamics and noise parameters based on transformation of the model into the Fitts' law form. In the field of Human-Computer Interaction and Ergonomics, the slope and intercept of Fitts' law have been used as an index of human motor performance because they represent the relationship between speed and accuracy. Our research results are significant in providing a first step to determining a specific relationship between the motor performance and arm dynamics.
Harris and Wolpert^[@CR20]^ and Tanaka *et al*.^[@CR21]^ showed that the SAT profile drawn by Fitts' law can be reproduced using their respectively proposed optimisation models. These were important results showing that Fitts' law for one-dimensional reaching movements can be reproduced based on single-joint arm dynamics. However, the multi-joint arm dynamics used in the more general arm movements performed by humans on a daily basis can cause nonlinear interference torques such as Coriolis and centrifugal forces between the upper arm and the forearm. As such the nonlinear interference torques do not occur under single-joint arm dynamics, it is important to investigate their effect on the SAT profile based on multi-joint arm dynamics. Our model, which is based on multi-joint arm dynamics, considers both Coriolis and centrifugal forces and therefore should be able to reproduce actual human arm movement more accurately than a model that does not consider them. To verify this, we compared applications of the model with and without Coriolis and centrifugal forces. An initial comparison using the data measured in the experiment described in Methods revealed an insignificant amount of difference between the two cases (Supplementary Fig. [S14](#MOESM1){ref-type="media"}). This occurred because the inertial and viscous forces dominated the Coriolis and centrifugal forces during this task and because the Coriolis and centrifugal forces in the shoulder joint cancelled each other in the Front and Back directions (Supplementary Fig. [S15](#MOESM1){ref-type="media"}). We then conducted an experiment with different starting and target postures to analyse the effects of moving rapidly over a longer distance. Supplementary Fig. [S16](#MOESM1){ref-type="media"} shows the results of the comparison for each subject. For five of the six subjects, the trial-averaged values of the actual endpoint error were better represented using the model with Coriolis and centrifugal forces, which also produced results closer the actual distributions of endpoint errors. These results demonstrate the importance of considering Coriolis and centrifugal forces in modelling very rapid reaching movements over long distances and highlight one of the advantages of our model, namely, that it considers the nonlinearity of arm movement.
Our model is based on the following assumptions and approximations: (1) feedforward control is assumed; (2) it is assumed that movement of the upper arm and forearm in the horizontal plane can be modelled by the dynamics of a two-link planar manipulator; (3) trajectory invariance in which the trajectory does not change within a certain range of movement duration is assumed; (4) the viscous force $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\frac{1}{D}B\dot{\tilde{\theta }}(s)$$\end{document}$ is approximated by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\frac{1}{{D}^{2}}\lambda B\dot{\tilde{\theta }}(s)$$\end{document}$ within a certain range of movement duration (Eq. ([6](#Equ6){ref-type=""}); (5) the presence of signal-dependent noise is assumed (Eq. ([7](#Equ7){ref-type=""})); (6) a first-order Taylor approximation of the forward model around $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\tau }_{i}^{{\rm{n}}oise}\mathrm{=0}$$\end{document}$ is applied (Supplementary Equation [2](#MOESM1){ref-type="media"}); (7) it is assumed that the effect of the partial derivatives of Coriolis and centrifugal forces are negligible (Supplementary Equation [3](#MOESM1){ref-type="media"}); and (8) a linear approximation is applied to formulate Fitts' law (Eq. ([18](#Equ18){ref-type=""})). These assumptions and approximations give several limitations to the model. For example, the effects of feedback cannot be considered because of assumption (1). Moreover, (3), (4), (6), (7) and (8) lead to modelling errors. It is also necessary to discuss the validity of the trajectory invariance assumption in (3). Whether or not time-normalisation is a valid must be confirmed in terms of how much the actually measured trajectory changes over the duration of movement. Supplementary Fig. [S1](#MOESM1){ref-type="media"} shows the measured hand paths of all subjects when moving a distance of 15 cm from 0.30 to 0.70 s. Although this provides a qualitative description, it can be considered to indicate that there is no problem with the changes even if trajectory invariance is assumed. This assumption is approximately valid to within a certain range of movement speeds that are neither too slow nor too fast. It should be noted that the assumption does not hold for very rapid or slow movement and, therefore, that finding a method that can be formulated without using time-normalisation should be undertaken as a future task. Another issue to be addressed in the future is the lack of a sophisticated understanding of motor impedance sources such as viscosity and stiffness. In this study, we did not consider joint stiffness because doing so would complicate the model. Given the numerous studies showing that stiffness is related to the SAT^[@CR26]--[@CR28]^, it will be important to incorporate stiffness into future versions of the model. Despite the limitations noted above, the error between the model and the experimental results was small, and we can conclude that the model can successfully explain the endpoint error occurring in actual movement.
The use of signal-dependent noise made it possible to represent the SATs---a breakthrough in the study of motor control. This breakthrough should be supplemented by further findings obtained, for example, by considering the constant and temporal noises in the model, as van Beers *et al*.^[@CR22]^ did; and considering how to estimate the noise parameter *k* using data other than hand endpoint error (e.g., a method that can estimate using EMG signals). We are also interested in studying human motor control strategies that achieve high spatial accuracy even with very rapid movement. It is speculated that nonlinear interference forces such as Coriolis and centrifugal forces are effectively used in such strategies, but this needs to be studied in detail.
Finally, in this study we investigated only arm-reaching movement. It will be possible to investigate other types of movement, such as eye and feet movement, have not, but changing the approach, for example, from arm to foot dynamics. The relationships derived in this paper can therefore be generally applied to any type of movement, which will be an important approach in future research.
Methods {#Sec11}
=======
We conducted behavioural experiments to examine to what degree the model endpoint error can represent the actual endpoint error and how similar the model is to Fitts' law when it is applied to the Fitts' law axis.
Subjects {#Sec12}
--------
Eleven healthy young adults (11 males, age range 21--24 y) participated. All subjects were right-handed according to the Edinburgh handedness test score (score range 64.7--100%). Informed consent agreements were obtained from all subjects, and the study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Nagaoka University of Technology.
Apparatus {#Sec13}
---------
The experimental setup is illustrated in Fig. [4A](#Fig4){ref-type="fig"}. The subjects sat on the chair located in front of the table and display. Each subject placed his right arm on the air-sled on the table to reduce the friction between their arm and the table, with the hand position set to match the tip of the handle of the air-sled. The air-sled and forearm were then fixed with stretch tape and the chair height was adjusted so that the table and subject's arm were oriented in a parallel manner. Each subject's shoulder was restrained by seat belts, restricting movement of the arm to the horizontal plane. The chair was then moved toward the table until the subject's chest just touched the table. Infrared markers were affixed to the shoulder and elbow joints and the hand, and the three marker positions were measured at a sampling frequency of 500 Hz using a three-dimensional optical position measurement digitiser (Optotrak Certus, Northern Digital Inc., Waterloo, Canada). The measured current hand positions were projected onto a display (PDP-504P, Pioneer, Tokyo, Japan) installed in front of the subject. A starting point (a circle with a radius of 10 mm) and a target (a circle with a radius of 2 mm) were also shown on the screen, and the subjects performed the experimental task while watching the display.Figure 4Experimental design. (**A**) Top view of experimental setup, showing definitions of joint angles *θ*~1~ and *θ*~2~ and *x*- and *y*-coordinates. (**B**) Display positions of all four targets with respect to central circle. (**C**) Examples of screen patterns of starting and target positions and experimental protocols. (**D**) Process for estimating signal-dependent noise parameters *k*~1~ and *k*~2~. To evaluate the degree of coincidence between pairs of ellipses, we defined the Ellipse Difference Index (EDI), which calculates the distances between four corresponding points (**A**--**D**) on the respective ellipses. The sum of the EDI over all conditions of movement duration and direction was set as the total EDI, with the values of *k*~1~ and *k*~2~ at the minimum total EDI selected as the signal-dependent noises.
Task {#Sec14}
----
Experimental tasks used in the study of SAT primarily fall into two categories: spatially constrained movement tasks, and temporally constrained movement tasks (details are given in^[@CR4]^).
In this study, we used a temporally constrained movement task based on our assumption of pre-programmed and rapid movements. The task involved setting a target movement duration *D* and movement distance *A* and then measuring the hand endpoint error or variability *W* of the corresponding movement. The subjects were required to reach the target as accurately as possible within the instructed range of movement duration.
Specifically, as shown in Fig. [4A](#Fig4){ref-type="fig"}, the task involved using the shoulder and elbow joints to engage in arm reaching movements in the horizontal plane toward the target, a circle with a radius of 2 mm, located 15 cm away from a 10-mm starting circle displayed in the centre of the screen. The initial shoulder position was defined as the origin of a coordinate system in which the positive *x*- and *y*-axes were to the right and in front of the subject, respectively. The starting position was defined as a shoulder angle (*θ*~1~) of 45° and an elbow angle (*θ*~2~) of 100°, and therefore the starting position relative to the right shoulder differed by subject: on average, it was −0.072 m toward the *x*-axis and 0.390 m toward the *y*-axis with respect to the initial shoulder position.
The targets were located at four directions---front (90°), back (270°), left (180°), and right (0°)---relative to the starting position (Fig. [4B](#Fig4){ref-type="fig"}). Movement was measured over the target movement duration intervals 0.30--0.40, 0.40--0.50, 0.50--0.60 and 0.60--0.70 s, respectively. As the distance from the starting point to the target was fixed, the movement speed was controlled. After the end of each trial, it was determined whether the measured movement duration was within the set duration range and whether the movement was executed ballistically. If both conditions were satisfied, the trial was regarded as a success; if neither was satisfied, it was regarded as a failure. Corresponding feedback was displayed to the subject on the upper right of the screen as follows: *failure* (*too fast*), *failure* (*too slow*), or *success* (Fig. [4A](#Fig4){ref-type="fig"}).
Figure [4C](#Fig4){ref-type="fig"} shows examples of the screen for each target pattern; the four target patterns were presented pseudorandomly. In the example shown, the pattern started from the front and, if the trial was a success, it continued backward, leftward and then rightward. The conditions of target movement duration were also applied pseudorandomly; in the example, the first interval was 0.60--0.70 s and was repeated five times until success was achieved in all four directions. If success was achieved over 4 directions × 5 rounds = 20 trials, the next movement duration was applied. If success was achieved at all movement duration conditions, the session was concluded. In all, four sessions were performed for each subject. As it was necessary for each subject to succeed at 80 trials per session, the experiment was continued until a total of 320 successful trials had been conducted.
Each subject participated in a preparatory experiment involving the same conditions as the main experiment but with half the number of sessions, with the goal of familiarising the subjects with the experimental setup and imparting a sense of target movement duration. The data measured in the preparatory experiment were not used in the subsequent analysis; in any event, no apparent learning effects were observed in the main experiments as a result of the preparatory sessions.
Data analysis {#Sec15}
-------------
### Observed endpoint error and movement duration {#Sec16}
Our methods for pre-processing the positional data and determining the observed endpoint error and movement duration were carried out as followed. The acquired positional data were low-pass filtered using a third-order, zero phase-lag Butterworth filter with a cut-off frequency of 10 Hz. The start and end of each movement were determined based on the tangential velocity, which was calculated using numerical differentiation. The start of a movement was defined as the point following the start cue at which the tangential velocity first exceeded the threshold velocity, defined as 5% of the peak value of the tangential velocity. The end of a movement was defined as the point prior to the end cue at which the tangential velocity fell below the threshold. The time from movement start to end was then defined as the observed movement duration, *D*. The required conditions of a successful trial were therefore (1) that the observed movement duration fell within the target range and (2) that there was only one peak in the tangential velocity between the start and end of the movement. The successful trials were then analysed to determine the corresponding Euclidean distances from the centre of the target to the endpoint, which was defined as the observed endpoint error, *W* ^observed^.
### Model endpoint error {#Sec17}
To obtain the model endpoint error, *W* ^model^, the following model parameters were used: the kinematic parameters such as the time-normalised desired angular position, velocity, and acceleration for each joint ($\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\tilde{\theta }}_{i}(s)$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\dot{\tilde{\theta }}}_{i}(s)$$\end{document}$, and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\ddot{\tilde{\theta }}}_{i}(s)$$\end{document}$), the physical parameters for each link or joint (*L*~*i*~, *S*~*i*~, *m*~*i*~, *I*~*i*~, and *B*~*ij*~), and the noise parameter for each joint (*k*~*i*~). We describe below how we estimated them.
Desired trajectories. We used the joint angular trajectories averaged over trials as the theoretically desired trajectory. The joint angular trajectory was first calculated using the measured positional data and subject's arm length by inverse kinematics and resampled temporally at the mean movement duration for each trial. The trial-averaged trajectories were then calculated for each subject and direction (for each target movement duration was also calculated as well). The trial-averaged joint angular velocities and acceleration were also calculated and time-normalised in terms of *D* to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\dot{\tilde{\theta }}(s)$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\ddot{\tilde{\theta }}(s)$$\end{document}$, respectively.
Physical parameters. The upper arm and forearm lengths, *L*~1~ and *L*~2~, respectively, were obtained from the distances between the markers on the shoulder, elbow, and hand, which were measured using the position digitiser before starting the experiment. The distance from the joint to the centre of mass, *S*~*i*~, mass, *m*~*i*~, and moment of inertia around the joint, *I*~*i*~, were estimated based on their respective proportional relationships to arm length^[@CR19],[@CR24],[@CR29]^. An adjusted forearm mass, *m*~2~, was obtained by adding the 0.740-kg mass of the air-sled and stretch tapes to the forearm mass estimated using the above method. The viscosity coefficient, *B*~*ij*~, was estimated using the method applied in^[@CR19]^ based on the approximate relationship between joint torque and viscosity coefficient during static force control, as measured by^[@CR30]^. The resulting physical parameters of the subject's arm are listed in Table [2](#Tab2){ref-type="table"}. Note that the viscosity coefficients shown in the table were obtained using the trial-averaged joint angular trajectory for each direction. For further details on how to calculate these physical parameters, refer to Supplementary Equation [4](#MOESM1){ref-type="media"}.Table 2Physical parameters of the arm and signal-dependent noise parameters.SubjectABCDEFGHIJKMean ± SD\**L*~1~ (m)0.3020.2700.2640.2610.2660.3080.2740.2780.2960.3030.2820.282 ± 0.017*L*~2~ (m)0.3250.3150.3350.3350.3120.3500.3160.3190.3450.3600.3320.331 ± 0.016*S*~1~ (m)0.1140.1010.0980.0970.0990.1160.1020.1040.1110.1140.1060.106 ± 0.007*S*~2~ (m)0.1590.1550.1640.1640.1530.1700.1550.1560.1680.1750.1620.162 ± 0.007*m*~1~ (kg)1.5061.3321.2991.2831.3101.5381.3541.3751.4731.5111.3971.398 ± 0.093*m*~2~ (kg)1.7941.7641.8241.8241.7551.8681.7671.7761.8531.8981.8151.813 ± 0.046*I*~1~ (kg · m^2^)0.0300.0210.0190.0190.0200.0320.0220.0230.0280.0300.0240.024 ± 0.005*I*~2~ (kg · m^2^)0.0400.0370.0430.0430.0360.0490.0370.0380.0470.0530.0420.042 ± 0.006**Front***B*~11~ (kg · m^2^/s)0.6980.6820.6850.6840.6840.7080.6810.6860.6930.6980.6940.690 ± 0.009*B*~12~ (kg · m^2^/s)0.1800.1810.1810.1810.1800.1830.1810.1820.1820.1820.1790.181 ± 0.001*B*~21~ (kg · m^2^/s)0.1800.1810.1810.1810.1800.1830.1810.1820.1820.1820.1790.181 ± 0.001*B*~22~ (kg · m^2^/s)0.7840.7890.7890.7910.7870.7990.7900.7940.7950.7940.7790.790 ± 0.006**Back***B*~11~ (kg · m^2^/s)0.6800.6680.6720.6720.6710.6880.6680.6670.6710.6810.6790.674 ± 0.007*B*~12~ (kg · m^2^/s)0.1860.1850.1860.1870.1850.1900.1850.1850.1870.1880.1860.186 ± 0.002*B*~21~ (kg · m^2^/s)0.1860.1850.1860.1870.1850.1900.1850.1850.1870.1880.1860.186 ± 0.002*B*~22~ (kg · m^2^/s)0.8130.8080.8160.8180.8110.8330.8080.8100.8210.8230.8130.816 ± 0.007**Left***B*~11~ (kg · m^2^/s)0.7470.7290.7330.7230.7230.7660.7340.7330.7530.7590.7420.740 ± 0.015*B*~12~ (kg · m^2^/s)0.1800.1800.1810.1800.1800.1810.1790.1800.1810.1810.1820.180 ± 0.001*B*~21~ (kg · m^2^/s)0.1800.1800.1810.1800.1800.1810.1790.1800.1810.1810.1820.180 ± 0.001*B*~22~ (kg · m^2^/s)0.7840.7840.7890.7870.7840.7890.7790.7830.7910.7920.7960.787 ± 0.005**Right***B*~11~ (kg · m^2^/s)0.7220.7090.7040.7040.7030.7370.7090.7060.7200.7270.7200.715 ± 0.011*B*~12~ (kg · m^2^/s)0.1810.1800.1810.1810.1810.1810.1800.1800.1810.1820.1820.181 ± 0.001*B*~21~ (kg · m^2^/s)0.1810.1800.1810.1810.1810.1810.1800.1800.1810.1820.1820.181 ± 0.001*B*~22~ (kg · m^2^/s)0.7880.7870.7920.7900.7870.7900.7830.7860.7900.7950.7970.790 ± 0.004*k*~1~0.5590.5370.5400.1710.6060.6790.6670.3000.2350.6010.5710.497 ± 0.177*k*~2~0.6670.6560.9290.7370.6540.9070.7320.8900.7280.7500.7150.760 ± 0.101\*SD: standard deviation across subjects.
Noise parameters. The parameter determining the magnitude of signal-dependent noise, *k*~*i*~, was estimated using the following simulation procedure:The joint torques were calculated using the mean trajectories and the physical parameters of the arm.Signal-dependent noise was generated artificially using an arbitrary value of *k*~*i*~ and added to the joint torque. The calculation applied Eq. ([7](#Equ7){ref-type=""}) with *z*~*i*~(*t*) set as a random variable with a standard normal distribution.The endpoint errors were obtained by transforming the noise-added torque from the joint to the task space using the forward dynamics and kinematics models.Step 1--3 were repeated 100 times for each pairing of *k*~1~ and *k*~2~ to obtain 100 simulated endpoints. We then used a grid search algorithm to find the value of *k*~*i*~ corresponding to the smallest difference between the confidence ellipses of the simulated and observed endpoints in terms of the Ellipse Difference Index (EDI), the sum of distances between four corresponding points on two ellipses (see Fig. [4D](#Fig4){ref-type="fig"}). The total EDI was defined as the summed value of the EDI over all movement duration and directions.
The noise parameters of each subject obtained using this method are listed at the bottom of Table [2](#Tab2){ref-type="table"}.
Supplementary information
=========================
{#Sec18}
Supplementary Information
**Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
=========================
is available for this paper at 10.1038/s41598-019-56016-7.
This work was partly supported by JSPS KAKENHI Grant Numbers 15K12597 and 18H04109, Nakatomi Foundation, and Nagaoka University of Technology Presidential Research Grant. We would like to thank Editage ([www.editage.jp](http://www.editage.jp)) for English language editing.
H.S., H.I. and Y.W. conceived the study and M.T., I.N. and Y.W. designed the experiment. M.T. developed the experimental setup, made the experiments and analysed the results with advice from T.S., I.N. and Y.W. M.T. wrote the manuscript. All authors read and approved the final manuscript.
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
The authors declare no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
The radical coupling reaction,[@cit1] wherein the nearly zero activation energy[@cit2] of connecting two distinct odd-electron partners enables a strong tendency to generate a new chemical bond, often furnishes an efficient and direct synthetic pathway to molecules with high functional group tolerance. Therefore, the development of compatible strategies to produce two types of radical species and facilitate the coupling reaction has attracted much attention from chemists over the last few decades.[@cit1]--[@cit3] In 2011, the MacMillan group made a significant breakthrough[@cit4] in which such a transformation was realized *via* visible-light-driven photoredox catalysis.[@cit5] This convenient and sustainable platform promptly inspired a number of elegant efforts focusing on radical coupling.[@cit6] However, the enantioselective manifold still remains underdeveloped given that few examples have been established.[@cit7] This lack of development is mainly due to the high reactivity that results in an elusive precise absolute stereocontrol of the reaction. In 2015, Ooi and co-workers[@cit7a] reported the first example with excellent enantioselectivity, wherein asymmetric α-coupling of *N*-arylaminomethanes with aldimines through cooperative catalysis[@cit7]--[@cit10] of an Ir-centered photosensitizer and an ionic Brønsted acid led to chiral 1,2-diamine derivatives featuring a tertiary carbon stereocenter. Almost at the same time, the Meggers group described an asymmetric radical coupling of tertiary amines with 2-trifluoroacetyl imidazoles catalyzed by a chiral iridium complex.[@cit7b] They also developed a cooperative Ru-centered photosensitizer and Rh-based chiral Lewis acid catalyst for the cross-coupling of 2-acyl imidazoles with α-silylamines.[@cit7d] In their elegant work, α-aminoalkyl moieties were successfully introduced onto the more challenging quaternary carbon stereogenic center. Inarguably, the judicious use of imidazole as the substituent of ketones is crucial for the enantioselectivity, as its N atom was shown to enhance the ability of stereocontrol by interacting with the chiral Lewis acid. Although the groups of Xiao,[@cit6f] Shah[@cit6g] and Yang[@cit6h] have demonstrated the viability of radical coupling of various readily accessible ketones with different substrates in a racemic manner, an enantioselective variety still constitutes a formidable challenge.
In the putative reaction mechanism of the radical-coupling transformations, ketones always experience a single-electron reduction to generate the corresponding ketyl variants.[@cit6],[@cit7] In 2013, the Knowles group[@cit9f] revealed the strong basicity of ketyls and their capability of forming neutral ketyl radicals with a chiral Brønsted acid which could provide efficient enantioselective control for the coupling reaction with hydrazones. This prominent H-bonding catalytic strategy recently inspired us to accomplish a highly enantioselective photoreduction of 1,2-diketones.[@cit10c] Accordingly, we speculated that such an approach would offer the possibility to address the desired radical coupling of undecorated ketones. If so, it should be a promising and general strategy as it would allow diverse radical species to connect with ketones, thus opening a fruitful avenue for the synthesis of the important chiral tertiary alcohols. Here, we report the development of a redox-neutral, enantioselective, radical coupling of *N*-aryl glycines with activated ketones, including acyclic 1,2-diketones and cyclic isatins ([Scheme 1](#sch1){ref-type="fig"}). The association of a dicyanopyrazine-derived chromophore (DPZ) as the photoredox catalyst and a 1,1′-spirobiindane-7,7′-diol (SPINOL)-based chiral phosphoric acid (CPA) as the H-bonding catalyst was demonstrated as being a workable catalytic system. Two series of chiral 1,2-amino tertiary alcohols that are significant structural scaffolds in synthetic and medicinal chemistry[@cit11] were prepared in high yields with good to excellent enantioselectivities.
{#sch1}
Results and discussion
======================
At the onset of our study, we sought to explore the coupling of ketyls derived from 1,2-diketones with α-amino radicals to provide the valuable enantioenriched 1,2-amino tertiary alcohols,[@cit11a] for which no catalytic asymmetric synthesis has yet been developed. Our developed metal-free DPZ[@cit7g],[@cit10] was selected as the photosensitizer since 1,2-diketones (*e.g.* benzil, *E*red1/2 = --1.169 V and --1.251 V *vs.* a saturated calomel electrode (SCE) in CH~3~CN)[@cit10c] could be directly reduced by DPZ˙^--^ (*E*red1/2 = --1.45 V *vs.* SCE in CH~2~Cl~2~) when the transformation experiences a reductive quenching cycle ([Scheme 2](#sch2){ref-type="fig"}). Furthermore, we intend to use *N*-aryl glycines (*e.g. N*-phenyl glycine, carboxylate ion, *E*~p~ = +0.52 \[for the N moiety\] and +1.09 V \[for the COO^--^ moiety\] *vs.* Ag/AgCl in CH~3~CN) as the precursors of α-amino radicals given the prominent ability to generate the radical species[@cit12]*via* single-electron oxidative decarboxylation by the visible-light-activated DPZ, *i.e.* \*DPZ (*E*^t^(S\*/S˙^--^) = +0.91 V *vs.* in CH~2~Cl~2~).[@cit7g],[@cit10d],[@cit10e] More importantly, the formed α-amino radicals contain a H-bonding donor (N--H) which will be more nucleophilic *via* a possible interaction with the H-bonding acceptor (X = N, O, *etc*.) of a chiral bifunctional catalyst, thus facilitating coupling with the electrophilic α-ketone radicals.[@cit9s],[@cit13] Of note, the use of tertiary amines was susceptible to reduce 1,2-diketones to α-hydroxy ketones under photoredox catalysis.[@cit10c],[@cit14]
{#sch2}
To this end, we began the investigation with *N*-phenyl glycine **1a** and benzil **2a** in the presence of DPZ as the photoredox catalyst ([Table 1](#tab1){ref-type="table"}).[@cit10c]--[@cit10e] The initial investigation showed that the transformation furnished racemic product **3a** in 62% yield after 12 h when only using 0.5 mol% DPZ at 25 °C, indicating the feasibility of the reaction and the high reactivity (entry 1). Upon examining a range of chiral H-bonding catalysts and the reaction parameters (see Table S1 in the ESI[†](#fn1){ref-type="fn"}), we observed that the reaction conducted in CPME at 10 °C for 36 h in the presence of 1.5 mol% DPZ and 10 mol% chiral SPINOL-CPA[@cit15]**C1** and with the combination of 30 mol% TBPB, 0.5 equiv. Na~2~S~2~O~4~ and 25 mg 5 Å MS as additives affords the desired chiral product **3a** in 78% yield with 93% ee (entry 2). Other SPINOL-CPAs (*e.g.*, **C2** and **C3**) with distinct substituents at the 6,6′-positions presented **3a** with lower ee values (entries 3--4). \[Ru(bpy)~3~\]Cl~2~ and Rose Bengal as plausible photoredox catalysts were evaluated (entries 5--6), but both the yield and enantioselectivity were decreased. Each of the three additives, which could regulate the strength of H-bonding interaction by exploiting the salt effect (*i.e.* TBPB[@cit10d] and Na~2~S~2~O~4~) or diminishing the moisture of the reaction system (*i.e.* molecular sieves[@cit7g]), was found to slightly affect the enantioselectivity to the same degree (entries 7--9). The results reveal that all of the additives jointly exerted a positive influence on the stereocontrol of **C1**. The control experiments confirmed that DPZ, visible light, and an oxygen-free environment are indispensable for the transformation to occur (entries 10--12). Note that the reaction performed under the standard conditions but in the absence of catalyst **C1** still produced *rac*-**3a** in 53% yield (entry 13), suggesting the existence of a considerable competitive achiral background reaction.
###### Optimization of the reaction conditions[^*a*^](#tab1fna){ref-type="fn"}
{#ugr1}
------------------------------- -------------------------------------------------------------------------------- ----------------------------------------- ------
1 0.5 mol% DPZ in CH~2~Cl~2~ at 25 °C and without **C1** and any additives, 12 h 62 N.A.
2 None 78 93
3 **C2** instead of **C1** 74 76
4 **C3** instead of **C1** 73 67
5 Ru(bpy)~3~Cl~2~·6H~2~O instead of DPZ 58 89
6 Rose Bengal instead of DPZ 69 91
7 No TBPB 76 89
8 No Na~2~S~2~O~4~ 75 90
9 No 5 Å MS 76 90
10 No DPZ 35 88
11 No light 0[^*d*^](#tab1fnd){ref-type="table-fn"} N.A.
12 Under air 0[^*e*^](#tab1fne){ref-type="table-fn"} N.A.
13 No **C1** 53 N.A.
^*a*^Reaction conditions: **1a** (0.075 mmol), **2a** (0.05 mmol), degassed CPME (1.0 mL), 10 °C, irradiation with blue LED (3 W, 450 nm), 36 h.
^*b*^Yield of the isolated product.
^*c*^Determined by HPLC analysis on a chiral stationary phase.
^*d*^No reaction was detected.
^*e*^ **1a** was consumed, but no **3a** was obtained. TBPB = tetra-*n*-butylphosphonium bromide. CPME = cyclopentyl methyl ether. N.A. = not available.
With the optimal reaction conditions in hand, the scope of this H-bonding catalysis-enabled asymmetric radical coupling strategy was examined ([Scheme 3](#sch3){ref-type="fig"}). The reactions of **2a** with a variety of *N*-aryl glycines **1** furnished adducts **3a--g** in 78 to 89% yields with 90 to 93% ees within 36 h. Electron-deficient or electron-donating substituents at the *para*- and *meta*-positions of the aryl ring presented similar reactivities and enantioselectivities. An attempt at performing the reaction of **1a** with **2a** in a 1.0 mmol scale presented a similar reactivity and enantioselectivity as **3a**, with 87% yield with 93% ee achieved after 48 h (footnote *a*). With respect to symmetric 1,2-diketones, the reaction tolerated a wide range of aryl substituents regardless of their electronic properties and substitution patterns, and corresponding products **3h--q** were obtained in 61 to 89% yields with 84 to 97% ees. Based on the persistent radical effect,[@cit13] the slightly lower enantioselectivity for 1,2-diketones (**3h--j**) with electron-withdrawing substituents is likely due to the stronger racemic background reaction, as the higher stability of these ketyl intermediates would facilitate a coupling with the unstable α-amino radical. For 3-methyl-1-phenyl-1,2-butanedione as a representative of unsymmetrical 1,2-diketones, product **3r** was obtained in only 34% ee. However, the modified reaction conditions, that are **C2** as the chiral catalyst and 4 Å MS as an additive in CPME at --5 °C, could furnish **3r** in 78% ee. The stereochemistry of these adducts was assigned based on the structure of **3q**, as solved by single crystal X-ray diffraction.[@cit16]
{#sch3}
The promising results inspired us to further evaluate this catalytic strategy for isatins, a representative cyclic ketone, to first construct the important chiral 3-hydroxy-3-aminoalkylindolin-2-one derivatives in a direct manner.[@cit11b]--[@cit11g] As depicted in [Scheme 4](#sch4){ref-type="fig"}, under the same catalysis platform but with modified reaction conditions (1.0 mol% DPZ, 20 mol% SPINOL-CPA **C3** in THF at 10 °C), the transformations of *N*-aryl glycine **1h** with *N*-Boc-substituted isatins **4** were complete within 36 h, providing the desired adducts **5a--l** with diverse substituents on the aromatic ring of the isatins in 73 to 91% yields with 85 to 94% ees. It was observed that the substituent group at the 4-position (*e.g.*, **5b--c**) presented a slightly decreased enantioselectivity, probably owing to the steric hindrance.
{#sch4}
The two series of enantiomerically enriched 1,2-amino-alcohols featuring oxo-hetero-quaternary carbon stereocenters are direct precursors to many bioactive natural and non-natural compounds ([Scheme 1](#sch1){ref-type="fig"}). For example, the treatment of adduct **3g** derived from 1,2-diketone using TCCA readily cleaved the PMP *N*-protective group ([Scheme 5](#sch5){ref-type="fig"}). The resultant amine **6** was then transformed into chiral 1,2-imidazolyl tertiary alcohol **7** that possesses oral antifungal activity[@cit11a] in 65% yield over two steps with 92% ee. The transformation of adducts from isatins was also carried out. The Boc-protected product **8** was rapidly obtained in 98% yield by treating **5a** with (Boc)~2~O and DMAP. The replacement of 2-methyl-4-methoxyphenyl with Ts as the *N*-protective group was performed through a sequential process involving the use of TCCA for the deprotection and TsCl for the protection, furnishing product **9** in 72% yield and without diminishing the ee. The results clearly indicate that chiral products **5** are excellent synthetic intermediates for conveniently synthesizing the biologically important 3-hydroxy-3-aminoalkylindolin-2-one variants as shown in [Scheme 1](#sch1){ref-type="fig"}.[@cit11b]--[@cit11g]
{#sch5}
Based on the previous examples[@cit10],[@cit12] and the product structure, the reaction between *N*-aryl glycines and ketones that underwent a single electron transfer (SET) redox radical coupling process to form the products is reasonable. The Stern--Volmer experiments[@cit17] confirmed that the catalytic cycle was triggered from the reductive quenching of \*DPZ ([Scheme 2](#sch2){ref-type="fig"}). To better understand the role of the chiral CPA in asymmetric induction, the transformation of *N*-phenyl-*N*-methyl glycine **10** with benzil **2a** under the standard reaction conditions as shown in [Table 1](#tab1){ref-type="table"} was carried out, and adduct **11** was obtained in 35% yield with 15% ee ([Scheme 6](#sch6){ref-type="fig"}). The lower yield than that achieved with the transformation of *N*-phenyl glycine **1a** (entry 2, [Table 1](#tab1){ref-type="table"}) was due to the deteriorated chemoselectivity, as the reduced product of **2a**, *i.e.*, benzoin, was obtained in a considerable amount. Note that in the absence of catalyst **C1**, **3a** was also produced in a decreased yield in the reaction of **1a** with **2a** (entry 13, [Table 1](#tab1){ref-type="table"}). According to the persistent radical effect,[@cit13] the results suggest the existence of an interaction between N--H of **1a** as a H-bonding donor and P0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 1111111111111111111111111111111111 1111111111111111111111111111111111 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000 0000000000000000000000000000000000O of the chiral CPA as a H-bonding acceptor, thus increasing the nucleophilicity of the α-amino radicals. In this context, the chiral CPA should serve as a bifunctional catalyst to activate the reaction and provide stereocontrol for the new C--C bond formation,[@cit9s] for which a ternary transition state as shown in [Scheme 2](#sch2){ref-type="fig"} is plausible.
{#sch6}
Conclusions
===========
In summary, we developed an enantioselective radical coupling of *N*-aryl glycines with diverse activated ketones, including acyclic 1,2-diketones and cyclic isatins, *via* visible-light-driven cooperative photoredox and chiral H-bonding catalysis. A range of significant enantioenriched 1,2-amino alcohols that feature an oxo-hetero-quaternary carbon stereocenter were obtained in high yields and ees. This work robustly demonstrates the viability of H-bonding catalysis for the highly reactive radical-coupling of ketones by directly providing a stereocontrolled environment for ketones. We believe that this dual catalytic system can serve as a powerful tool to address a variety of radical coupling reactions of diverse ketones through flexibly selecting a photoredox catalyst and a H-bonding catalyst, thus providing a direct and productive approach to access various chiral tertiary alcohols. We also anticipate that this work will inspire the pursuit of novel enantioselective radical coupling for other oxidative substrates with feasible H-bonding acceptor moieties.
Conflicts of interest
=====================
There are no conflicts to declare.
Supplementary Material
======================
Supplementary information
######
Click here for additional data file.
Crystal structure data
######
Click here for additional data file.
Grants from the NSFC (21672052) and Provincial Innovation Scientists and Technicians Troop Construction Projects are gratefully acknowledged. We also appreciate Mr Yangyang Shen (ICIQ) for the constructive discussions.
[^1]: ‡The authors made equal contributions.
[^2]: †Electronic supplementary information (ESI) available. CCDC [1836991](1836991) and [1843336](1843336). For ESI and crystallographic data in CIF or other electronic format see DOI: [10.1039/c8sc02948b](10.1039/c8sc02948b)
| {
"pile_set_name": "PubMed Central"
} |
Lu N, Lv Q, Sun X, et al. Isoalantolactone restores the sensitivity of gram‐negative *Enterobacteriaceae* carrying MCR‐1 to carbapenems. J Cell Mol Med. 2020;24:2475--2483. 10.1111/jcmm.14936
Lu and Lv contributed equally to this work.
1. INTRODUCTION {#jcmm14936-sec-0001}
===============
Infections caused by multidrug‐resistant (MDR) gram‐negative pathogens have become one of the greatest threats to biohealth worldwide, leading to a huge challenge in clinical treatment.[1](#jcmm14936-bib-0001){ref-type="ref"} Multidrug‐resistant organisms such as *Klebsiella pneumoniae* and *E coli* have evolved multiple resistance mechanisms that limit treatment options.[2](#jcmm14936-bib-0002){ref-type="ref"} Polymyxins have good bactericidal effects against gram‐negative pathogens and are regarded as the last line of defence for the treatment of such bacterial pathogens.[3](#jcmm14936-bib-0003){ref-type="ref"}, [4](#jcmm14936-bib-0004){ref-type="ref"} Additionally, among these antibiotics, polymyxin B and colistin are cationic peptide antibiotics with lipophilic acyl side chains[5](#jcmm14936-bib-0005){ref-type="ref"} that were developed in the 1940s but fell into disfavour because of their high toxicity rates.[6](#jcmm14936-bib-0006){ref-type="ref"} The mechanism by which polymyxin kills gram‐negative pathogens relies on disrupting membrane permeability through polar and hydrophobic interactions. In these interactions, there is an electrostatic interaction between the positively charged residues of polymyxin and the negatively charged lipid A moiety of the lipopolysaccharide.[7](#jcmm14936-bib-0007){ref-type="ref"}, [8](#jcmm14936-bib-0008){ref-type="ref"} With the emergence of multidrug‐ and pandrug‐resistant gram‐negative pathogens in the 1990s, especially carbapenem‐resistant *Enterobacter*, we have to reactivate polymyxin in the absence of new antibacterials.[9](#jcmm14936-bib-0009){ref-type="ref"}, [10](#jcmm14936-bib-0010){ref-type="ref"}, [11](#jcmm14936-bib-0011){ref-type="ref"}
Since the first discovery of the transferable colistin resistance gene *mcr‐1* in China, more than 40 countries and regions around the world have detected such resistance genes.[12](#jcmm14936-bib-0012){ref-type="ref"} MCR‐1, a protein product of the *mcr‐1* gene, is predicted to be an integral membrane protein with the catalytic activity of phosphoethanolamine transferases. The MCR‐1 enzyme modifies the chemical structure of the lipid A moiety on bacterial LPS by the addition of phosphoethanolamine, which in turn reduces the binding affinity of LPS to colistin.[13](#jcmm14936-bib-0013){ref-type="ref"} The *mcr‐1* gene can be spread between different bacteria in various regions by plasmids or transposition, which not only poses a great threat to public health but also poses a great challenge to clinical anti‐infective treatment.[14](#jcmm14936-bib-0014){ref-type="ref"} The emergence of the transferable polymyxin resistance genes *mcr‐1* represents a new mechanism of bacterial resistance that greatly challenges the last line of defence for the treatment of multidrug‐resistant gram‐negative pathogens.[15](#jcmm14936-bib-0015){ref-type="ref"}
There is a need for an MCR inhibitor that works synergistically with polymyxin to treat infections caused by polymyxin B‐resistant *mcr‐1*‐positive *Enterobacteriaceae*. Isoalantolactone (IAL) has been shown to possess various pharmacological activities, including antitrypanosomal, anti‐apoptosis and antimicrobial activities.[16](#jcmm14936-bib-0016){ref-type="ref"}, [17](#jcmm14936-bib-0017){ref-type="ref"}, [18](#jcmm14936-bib-0018){ref-type="ref"} In addition, it has been reported that IAL has anticancer activity against several cancer cells, such as cervical squamous cell carcinoma, prostate cancer and gastric adenocarcinoma.[19](#jcmm14936-bib-0019){ref-type="ref"}, [20](#jcmm14936-bib-0020){ref-type="ref"} However, as a promising natural compound, the inhibitory effect of IAL against bacterial resistance enzymes has not been reported.
Here, we screened IAL as a new MCR‐1 inhibitor with the ability to enhance the sensitivity of *mcr‐1*‐positive *Enterobacteriaceae* to polymyxin B. Furthermore, the synergistic effect of IAL and polymyxin B was determined both in vivo and in vitro*.*
2. MATERIALS AND METHODS {#jcmm14936-sec-0002}
========================
2.1. Bacterial strains and reagents {#jcmm14936-sec-0003}
-----------------------------------
The strains used in this experiment were described in our preceding studies[21](#jcmm14936-bib-0021){ref-type="ref"}, [22](#jcmm14936-bib-0022){ref-type="ref"}, [23](#jcmm14936-bib-0023){ref-type="ref"} and are listed in Table [1](#jcmm14936-tbl-0001){ref-type="table"}. Isoalantolactone (IAL) was purchased from Chengdu Ruifensi Biotechnology Co. Ltd. Colistin, polymyxin B and penicillin were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China). Streptomycin sulphate, kanamycin sulphate, erythromycin and achromycin were purchased from Dalian Meilun Biotechnology Co Ltd. Stock solutions of IAL were prepared in dimethyl sulfoxide (Sigma‐Aldrich) at a concentration of 20 mg/mL.
######
MIC values of the different antibiotics and ISO combination therapy for each of the tested bacterial isolates
Species Source *mcr‐1* confirmation Antibiotics MIC (μg/mL) FIC index
----------------------------------------------- ------------------------------------------------------------------------------- ---------------------- ----------------- --------------- ------------- -------------------
*Escherichia coli* BL21(DE3) (pET28a‐*mcr‐1*) Laboratory strain carrying the *mcr‐1* gene from *Klebsiella pneumoniae* ZJ05 \+ Polymyxin B 10.00 ± 3.46 1.13 ± 0.54 **0.0.20 ± 0.08**
Colistin 16.00 ± 0.00 2.00 ± 0.00 **0.19 ± 0.00**
*E coli* ZJ478 Human intra‐abdominal fluid \+ Polymyxin B 16.00 ± 11.31 4.67 ± 2.49 **0.40 ± 0.12**
Colistin 8.00 ± 0.00 3.00 ± 1.41 **0.44 ± 0.18**
*E coli* DZ2‐12R Chicken cloacae \+ Polymyxin B 7.33 ± 4.27 4.67 ± 1.49 **0.41 ± 0.12**
Colistin 8.00 ± 0.00 3.00 ± 1.00 **0.44 ± 0.13**
Ciprofloxacin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Kanamycin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Erythromycin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Penicillin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Tetracycline 128.00 ± 0.00 256.00 ± 0.00 2.06 ± 0.00
*K pneumoniae* ZJ02 Remote tertiary care hospital \+ Polymyxin B 64.00 ± 0.00 3.60 ± 0.80 **0.13 ± 0.03**
Colistin 32.00 ± 0.00 6.00 ± 2.00 **0.25 ± 0.06**
Ciprofloxacin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Kanamycin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Erythromycin 256.00 ± 0.00c 256.00 ± 0.00 1.06 ± 0.00
Penicillin 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
Tetracycline 256.00 ± 0.00 256.00 ± 0.00 1.06 ± 0.00
*K pneumoniae* E831 Chicken cloacae \+ Polymyxin B 24.00 ± 8.00 6.00 ± 2.00 **0.38 ± 0.19**
Colistin 32.00 ± 0.00 6.00 ± 2.00 **0.25 ± 0.06**
*K pneumoniae* ZJ05 Remote tertiary care hospital \+ Polymyxin B 40.00 ± 13.86 8.00 ± 4.90 **0.25 ± 0.06**
Colistin 48.00 ± 16.00 3.00 ± 1.00 **0.13 ± 0.00**
*K pneumoniae* 13b5 Chicken cloacae \+ Polymyxin B 32.00 ± 0.00 8.00 ± 0.00 **0.31 ± 0.00**
Colistin 32.00 ± 0.00 4.00 ± 0.00 **0.19 ± 0.00**
*K pneumoniae* K7 The People\'s Hospital of Jilin Province − Polymyxin B 1.38 ± 0.75 1.50 ± 0.58 1.69 ± 1.60
Colistin 3.67 ± 3.79 1.00 ± 0.87 **0.40 ± 0.14**
*E coli* ATCC25922 ATCC − Polymyxin B 0.50 ± 0.00 0.50 ± 0.00 1.06 ± 0.00
colistin 0.50 ± 0.00 0.50 ± 0.00 1.06 ± 0.00
*E coli* BL21(DE3) (pET28a) Laboratory strain − Polymyxin B 0.75 ± 0.35 0.75 ± 0.35 1.06 ± 0.00
colistin 0.50 ± 0.00 0.38 ± 0.18 0.81 ± 0.35
The concentration of ISO was 32 μg/mL for all tested strains. All the MIC results were counted as no less than 3 times. The FICs lower than 0.5 was displayed in bold and recognized as synergistic effect.
John Wiley & Sons, Ltd
2.2. FIC values determination {#jcmm14936-sec-0004}
-----------------------------
The microbroth checkerboard dilution method recommended by the Nation Committee for Clinical Laboratory Standards (NCCLS)[24](#jcmm14936-bib-0024){ref-type="ref"}, [25](#jcmm14936-bib-0025){ref-type="ref"} was employed to evaluate the combined bactericidal effect of IAL and polymyxin B. The strains listed in Table [1](#jcmm14936-tbl-0001){ref-type="table"} were inoculated into 2 mL of LB medium overnight at 37°C with shaking at 200 rpm. The polymyxin B was diluted with LB medium for the antibiotic group without IAL, and the other group was combined with 32 μg/mL IAL. The results were observed after 24 hours of culture. The FIC values were determined as follows: FIC index = (FIC of polymyxin) + (FIC of IAL).[26](#jcmm14936-bib-0026){ref-type="ref"}
2.3. Growth curve determination {#jcmm14936-sec-0005}
-------------------------------
The tested strains *K pneumoniae* ZJ02, *E coli* DZ2‐12R or *E coli* BL21(DE3) (pET28a‐*mcr‐1*) were inoculated into a 2 mL of LB overnight and expanded into 100 mL of LB medium. When the OD value of the bacterial solution reached 0.3, the LB medium was quantitatively dispensed and supplemented with IAL at final concentrations of 0, 16, 32, 64 or 128 μg/mL. The cultures were grown at 37°C with shaking at 200 rpm, and a growth curve was plotted by monitoring the OD~600nm~ value at 0.5‐ to 1‐hour intervals.
2.4. Time‐killing assays {#jcmm14936-sec-0006}
------------------------
The bactericidal curve was used to evaluate the potential bactericidal effect of IAL combined with polymyxin B. For the tested strains *K pneumoniae* ZJ02, *E coli* BL21(DE3)(PET28a‐*mcr‐1*) and *E coli* DZ2‐12R, the concentrations of polymyxin B used in this study were 4, 2 and 2 μg/mL, respectively. The concentration of IAL was 32 μg/mL. The tested strains were treated with IAL, polymyxin B, a combination of IAL and polymyxin B, or were not treated (positive control). The bacterial solution of each sample was diluted with PBS, plated on TSB solid medium and grown for 0, 1, 3, 5 and 7 hours. Following cultivation in a 37°C incubator for 24 hours, the bacteria on the plate were counted to construct the time‐killing curve.
2.5. Combined disc test {#jcmm14936-sec-0007}
-----------------------
The bacterial solutions (MCR‐1‐positive *E coli* BL21(DE3) (pET28a‐*mcr‐1*), *E coli* DZ2‐12R and *K pneumoniae* ZJ02 or negative control bacteria, *E coli* ATCC 25922) were diluted 1:4000 and uniformly coated on the surface of LB solid medium containing various concentrations of IAL (0, 4 or 32 μg/mL). Then, a tablet containing polymyxin B was placed in the middle of the plate. The diameter of the inhibition zones was measured to evaluate the combination effect after being placed in a 37°C incubator for 24 hours.
2.6. Western blot assay {#jcmm14936-sec-0008}
-----------------------
The tested strains *K pneumoniae* ZJ02, *E coli* DZ2‐12R or *E coli* BL21 (DE3) (pET28a‐*mcr‐1*) were cocultured with various concentrations of IAL (0, 4 and 32 μg/mL) as described in the growth curve determination for 4 hours. Following centrifugation at 13 800 *g* for 5 min, the precipitate of each sample was resuspended in loading buffer, mixed, boiled at 100°C for 7 minutes and separated by SDS‐PAGE.[27](#jcmm14936-bib-0027){ref-type="ref"} Then, the protein was transferred to a PVDF membrane. After blocking, the mouse monoclonal antibody against MCR‐1 (1:2000, Laboratory preservation), goat antimouse HRP‐conjugated secondary antibodies (1:2000, Proteintech) were applied for the determination of MCR‐1 production as described in our previous study.[23](#jcmm14936-bib-0023){ref-type="ref"} Isocitrate dehydrogenase (ICDH) was used as an internal control according to a previous study.
2.7. Ethics statement {#jcmm14936-sec-0009}
---------------------
Six‐ to eight‐week‐old female Balb/c mice with an average bodyweight of 20 ± 2 g were purchased from the Experimental Animal Center of Jilin University (Changchun, Jilin, China). Animal experiments were approved by and conducted in accordance with the guidelines of the Animal Care and Use Committee of Jilin University.
2.8. Mouse thigh infection with *K pneumoniae* ZJ02 and *E coli* DZ2‐12R {#jcmm14936-sec-0010}
------------------------------------------------------------------------
Establishment of a model of mouse thigh muscle infection by experimental methods in the literature.[28](#jcmm14936-bib-0028){ref-type="ref"} In this experiment, *K pneumoniae* ZJ02 or *E coli* DZ2‐12R was injected intramuscularly into the thigh muscle with 2 × 10^7^ CFUs per mouse. After inoculation, the mice were divided randomly into four groups (n = 5): (a) solvent (50 μL of DMSO), control treatment group; (b) IAL (50 mg/kg of bodyweight), alone treatment group; (c) polymyxin B (20 mg/kg), alone treatment group; and (d) a combination treatment group with IAL (50 mg/kg) and polymyxin B (20 mg/kg). Following simultaneous subcutaneous administration and subsequent administration every 8 hours, the infected mice of each group were killed at 36 hours post‐infection. The muscle tissue was taken, ground evenly, diluted with PBS and plated to determine the bacterial burden in the thigh muscle of infected mice.
2.9. Survival analysis {#jcmm14936-sec-0011}
----------------------
Overnight culture of *E coli* DZ2‐12R was injected intraperitoneally into mice with 2 × 10^8^ CFUs per mouse. After inoculation, the mice were divided randomly into four groups (n = 10): (a) solvent (50 μL of DMSO), control treatment group; (b), IAL (50 mg/kg of bodyweight), alone treatment group; (c) polymyxin B (5 mg/kg), alone treatment group; and (d) a combination treatment group with IAL (50 mg/kg) and polymyxin B (5 mg/kg). Following simultaneous subcutaneous administration and subsequent administration every 8 hours for 72 hours, the survival of infected mice was observed for 1 week.
2.10. Lung tissue section experiment {#jcmm14936-sec-0012}
------------------------------------
Overnight culture of *K pneumoniae* ZJ02 was injected transnasally into mice with 5 × 10^7^ CFUs per mouse. After inoculation, the mice were divided randomly into four groups (n = 5): (a) solvent (50 μL of DMSO), control treatment group; (b) IAL (50 mg/kg of bodyweight), alone treatment group; (c) polymyxin B (5 mg/kg), alone treatment group; and (d) a combination treatment group with IAL (50 mg/kg) and polymyxin B (5 mg/kg). Following simultaneous subcutaneous administration and subsequent administration every 8 hours, the infected mice of each group were killed at 72 hours post‐infection. The lung was taken, immersed in 4% formalin, stained with haematoxylin and eosin and visualized under a microscope.
2.11. Statistical analysis {#jcmm14936-sec-0013}
--------------------------
The data are presented as the average plus the relative standard deviation and analysed with Student\'s *t* test. *P*‐value of \<.05 was considered significant. \*, *P* \< .05; \*\*, *P* \< .01.
3. RESULTS {#jcmm14936-sec-0014}
==========
3.1. IAL enhances *mcr*‐*1‐*positive *Enterobacteriaceae* sensitivity to carbapenems without affecting bacterial growth {#jcmm14936-sec-0015}
-----------------------------------------------------------------------------------------------------------------------
The checkerboard method is commonly used to screen inhibitors of resistance enzymes.[29](#jcmm14936-bib-0029){ref-type="ref"} Here, we found that 32 μg/mL IAL, combined with polymyxin B or colistin, had a significantly synergistic effect against the engineering strain *E coli* BL21(DE3) (pET28a‐*mcr‐1*) (Table [1](#jcmm14936-tbl-0001){ref-type="table"}). Consistent with this result, all FIC values were lower than 0.5 for the *mcr‐1*‐positive *E coli* strains (ZJ478 and DZ2‐12R) and *K pneumoniae* strains (ZJ02, E831, E831 and 13b5) following a combination treatment with IAL (32 μg/mL) and carbapenems (polymyxin B or colistin). The combination treatment with IAL and polymyxin B resulted in a 16‐fold reduction (from 64 to 4 μg/mL) in the MICs of *K pneumoniae* ZJ02, a clinical strain carrying *mcr‐1.* Although a synergistic effect was observed for the combination of IAL and colistin for *K pneumoniae* K7, the combined treatment with IAL and carbapenems (polymyxin B or colistin) showed no such synergistic effect for *mcr‐1*‐negative *E coli* strains (BL21(DE3) (pET28a) and ATCC25922) and *K pneumoniae* K7. Thus, these results indicated that IAL is an effective MCR‐1 inhibitor with the ability to restore the antibacterial activity of carbapenems (polymyxin B or colistin) against *mcr‐1*‐positive bacteria.
However, the treatment with IAL alone at concentrations varying from 16 to 128 μg/mL had no effect on the growth of the tested strains (*E coli* BL21(DE3) (pET28a‐*mcr‐1*), *E coli* DZ2‐12R and *K pneumoniae* ZJ02, shown in Figure [1](#jcmm14936-fig-0001){ref-type="fig"}A‐C, respectively), suggesting that no antibacterial activity existed at IAL concentrations lower than 128 μg/mL. Furthermore, the synergistic effect of IAL and carbapenems was also observed in the time‐killing analysis (Figure [1](#jcmm14936-fig-0001){ref-type="fig"}D,F). For the tested strains, neither IAL nor polymyxin B has no bactericidal effect. The combination of polymyxin and IAL led to a bactericidal effect for *E coli* BL21 (DE3) (pET28a‐*mcr‐1*) and significant inhibition of bacterial growth for *E coli* DZ2‐12R and *K pneumonia* ZJ02 within 7 hours (Figure [1](#jcmm14936-fig-0001){ref-type="fig"}D‐F). In agreement with the above results, IAL combined with polymyxin significantly increased the inhibition zone from 11.17 ± 0.24 mm to 16.73 ± 1.32 mm for the *mcr‐1*‐positive strains *E coli* DZ2‐12R and *K pneumoniae* ZJ02 but not the *mcr‐1*‐negative strain *E coli* ATCC25922 (Figures [2](#jcmm14936-fig-0002){ref-type="fig"} and [3](#jcmm14936-fig-0003){ref-type="fig"}). Taken together, our results established that IAL, as an effective MCR‐1 inhibitor with negative antibacterial activity, recovered the antibacterial activity of carbapenems against *mcr‐1*‐positive *Enterobacteriaceae*.
{#jcmm14936-fig-0001}
{#jcmm14936-fig-0002}
{ref-type="fig"} was measured with a ruler. A, MCR‐1‐positive bacteria (*Escherichia coli* BL21(DE3) (pET28a‐*mcr‐1*). B, MCR‐1‐positive bacteria *E coli* DZ2‐12R. C, MCR‐1‐positive bacteria *Klebsiella pneumoniae* ZJ02. D, MCR‐1‐negative bacteria (*E coli* ATCC 25 922). \*, *P* \< .05; \*\*, *P* \< .01](JCMM-24-2475-g003){#jcmm14936-fig-0003}
3.2. IAL had no inhibitory effect on MCR‐1 production {#jcmm14936-sec-0016}
-----------------------------------------------------
A direct inhibition of MCR‐1 activity or MCR‐1 production may contribute to the synergistic effect of IAL with carbapenems. Western blot analysis was further employed to evaluate whether IAL treatment inhibits MCR‐1 production.[30](#jcmm14936-bib-0030){ref-type="ref"} As shown in Figure [4](#jcmm14936-fig-0004){ref-type="fig"}, the treatment with IAL at the concentrations required for the synergistic effect had no visible influence on the production of MCR‐1 by the tested strains. Thus, our results suggested that IAL effectively inhibits MCR‐1 without affecting the production of this enzyme protein.
{#jcmm14936-fig-0004}
3.3. Combination therapy of IAL and polymyxin B synergistically inhibits the pathogenicity of *mcr‐1*‐positive bacteria in vivo {#jcmm14936-sec-0017}
-------------------------------------------------------------------------------------------------------------------------------
To detect the same synergistic bactericidal effect of IAL and polymyxin B for *K pneumoniae* ZJ02 and *E coli* DZ2‐12R in vivo, a model of mouse thigh muscle infection was established in this study[31](#jcmm14936-bib-0031){ref-type="ref"} by assessing the bacterial burden in the thigh 36 hours post‐infection. As shown in Figure [5](#jcmm14936-fig-0005){ref-type="fig"}A, treatment with IAL or polymyxin B alone did not result in an evident reduction in the bacterial burden in *E coli* DZ2‐12R‐infected mice. However, a combination therapy of IAL and polymyxin B significantly reduced the bacterial colonization in the thigh (Figure [5](#jcmm14936-fig-0005){ref-type="fig"}A). Although a significant difference between the polymyxin B treatment and solvent control was observed for the *K pneumoniae* ZJ02‐infected mice, the combined therapy resulted in a more than 1.5‐log~10~ reduction of the bacterial burden in infected mice when compared with the polymyxin B treatment (Figure [5](#jcmm14936-fig-0005){ref-type="fig"}B). Together, these results indicated that IAL combined with polymyxin B synergistically inhibited the bacterial burden in mice.
{#jcmm14936-fig-0005}
3.4. Protection rate experiment {#jcmm14936-sec-0018}
-------------------------------
To further evaluate this synergistic effect, the survival rate of *E coli* DZ2‐12R in intraperitoneally infected mice treated with IAL, polymyxin B or a combination of IAL and polymyxin B was monitored for 72 hours. As expected, all infected mice were dead after 72 hours. After infection, the mice were monitored for 3 days using the same treatment method, combined with the solvent control group and the IAL group alone, and the survival rate of the mice in the antibiotic group alone was 20% (Figure [6](#jcmm14936-fig-0006){ref-type="fig"}). The combined polymyxin B and IAL group had a survival rate of 60% (Figure [6](#jcmm14936-fig-0006){ref-type="fig"}), further confirming that the combination treatment has a better effect than the drug or antibiotic alone.
{#jcmm14936-fig-0006}
Finally, the pathological damage in the lung was assessed in a mouse model of pneumonia by *K pneumoniae* ZJ02. As shown in Figure [7](#jcmm14936-fig-0007){ref-type="fig"}, no pathological damage was observed for the uninfected mice under the microscope. In the infected mice, the alveolar cavity was filled with exuded cellulose, alveolar septum thickening and capillaries. Consistent with the above results, the symptoms of the drug‐treated group and the antibiotic‐treated group were slightly relieved (Figure [7](#jcmm14936-fig-0007){ref-type="fig"}). However, the cellulose exudation in the alveolar cavity was reduced, and the pulmonary interstitial was only infiltrated by a small amount of inflammatory cells, suggesting that the pathological damage was significantly reduced (Figure [7](#jcmm14936-fig-0007){ref-type="fig"}). Taken together, IAL combined with polymyxin B remarkably inhibited the pathogenicity of *mcr‐1*‐positive bacteria in mice.
{#jcmm14936-fig-0007}
4. DISCUSSION {#jcmm14936-sec-0019}
=============
The emergence and spread of drug‐resistant bacteria that have caused serious clinical infections is a matter of great concern to researchers. The prevalence of *mcr‐1*‐positive *K pneumoniae* and *E coli* has brought great challenges to clinical treatment, which calls for novel strategies or agents to fight these bacterial pathogens. Among these strategies, the identification of effective MCR‐1 inhibitors combined with antibiotics is a promising strategy to address this challenge. Here, we successfully screened the natural compound IAL as an effective agent with the capability of restoring the antibacterial activity of carbapenems against *mcr‐1*‐positive *Enterobacteriaceae*. These results suggest that IAL is a potential leading compound for the treatment of *mcr‐1*‐positive bacteria when combined with antibiotics.
Isoalantolactone (IAL), the main sesquiterpene lactone in Radix Inulae and other plants, is a frequently utilized herbal medicine. Isoalantolactone has various pharmacologic effects,[31](#jcmm14936-bib-0031){ref-type="ref"}, [32](#jcmm14936-bib-0032){ref-type="ref"} such as inhibiting inflammation, preventing proliferation and inducing apoptosis.[33](#jcmm14936-bib-0033){ref-type="ref"} Furthermore, this compound has been reported to significantly induce breast cancer cell apoptosis by activating the caspase cascade, cleaving poly (ADP‐ribose) polymerase.[34](#jcmm14936-bib-0034){ref-type="ref"} In addition, IAL was recently identified as selectively toxic to cancer cells.[18](#jcmm14936-bib-0018){ref-type="ref"}, [35](#jcmm14936-bib-0035){ref-type="ref"} These results suggested that IAL is a multiple biological compound, as a drug, used in clinical practice. However, IAL is not soluble in water, which limited its pharmacological activity in vitro/in vivo. Therefore, the molecular structure of IAL requires further modification.
In the present study, we confirmed the synergistic effect of polymyxin B and IAL on plasmid‐mediated *mcr‐1*‐positive *Enterobacteriaceae*, both in vitro and in vivo. From checkerboard microdilution assays, we found that IAL only has a synergistic effect on *mcr‐1*‐positive *Enterobacteriaceae* in combination with carbapenem antibiotics, but it has no effect when used in combination with other antibiotics. Moreover, IAL has no synergistic effect on *mcr‐1*‐negative isolates (Table [1](#jcmm14936-tbl-0001){ref-type="table"}). Additionally, IAL treatment did not affect the production of MCR‐1 in *mcr‐1*‐positive *Enterobacteriaceae* (Figure [4](#jcmm14936-fig-0004){ref-type="fig"})*.* This particular synergistic effect suggests that IAL may act directly on MCR‐1 and affect its activity. Furthermore, the combination of isoalantolactone and polymyxin B had the best synergistic effect on *Klebsiella pneumonia*, with a 16‐fold change in the MIC values (Table [1](#jcmm14936-tbl-0001){ref-type="table"}). Upon examination, compared with other MCR‐1 inhibitors, IAL has a great advantage with a good synergistic effect on polymyxin‐resistant gram‐negative bacteria both in vitro and in vivo. All the information described suggested that IAL possesses the possibility for development into a pharmaceutical preparation for clinical use.
Some clinical drugs are often limited in their use because of poor absorption in vivo. However, in this study, subcutaneous administration of IAL and polymyxin B achieved good therapeutic effects for the animal infection treatment experiments. Our data indicated that absorption may not affect the usage of IAL as a drug.
Taken together, our results suggest that the combination of IAL and carbapenems can be an option for the treatment of MCR‐1‐positive *Enterobacteriaceae*. As a promising natural compound, IAL has more pharmacological activities to be discovered.
CONFLICT OF INTEREST {#jcmm14936-sec-0021}
====================
The authors have no conflict of interest to declare.
AUTHOR CONTRIBUTIONS {#jcmm14936-sec-0022}
====================
LN, LQ, SX, ZY, ZP and WJ conceived and performed all the experiments. LN, LQ and WJ researched the data contributed to the statistical analysis and discussion. QJ contributed to the discussion and reviewed the manuscript. All authors reviewed the manuscript.
This work was supported by the National Key Technology R&D Program (No. 2018YFD0500300), the National Natural Science Foundation of China (81861138046), the Jilin Scientific and Technological Development Program under Grant 20190103147JH and the Fundamental Research Funds for the Central Universities.
DATA AVAILABILITY STATEMENT {#jcmm14936-sec-0024}
===========================
The data used to support the findings of this study are available from the corresponding author upon request.
| {
"pile_set_name": "PubMed Central"
} |
Cold storage has been the method of choice for liver preservation for \<50 years now. Hypothermia effectively protects the organ by reduced metabolism for distinct periods. However, the organ is exposed to hypoxic, hypoenergetic, and hyponutritional conditions. These are nonfunctional in the cold but unmasked upon reperfusion and trigger cellular signal pathways leading to hepatocellular injury and compromised graft recovery.^[@R1]^ Therefore, research in the last decades focused on methods of dynamic preservation, allowing for continuous supply of the allograft with oxygen and nutrients as well as removal of metabolic waste products. Machine perfusion of the liver can be performed under various conditions. The concept of controlled oxygenated rewarming (COR) is an end-ischemic allograft perfusion model, which utilizes the advantages of cold storage for transportation of the allograft and is meant to recondition the allograft in the accepting transplantation center before liver transplantation (LT). It is based on the observations that the abrupt temperature shift on warm reperfusion triggers mitochondrial dysfunction due to a progressive mitochondrial transition pore opening.^[@R2]--[@R4]^ Preceding studies have shown that slowly increasing the temperature from the cold provides a transient phase of cold to mid-thermic perfusion, which allows for a gentle restitution of mitochondrial function at limited workload and demonstrates excellent restitution of hepatocellular function. This alleviated the trigger for mitochondrial dysfunction upon normothermic reperfusion. The first clinical application of COR before liver transplantation was performed in 6 extended criteria donor liver allografts and demonstrated convincing clinical results^[@R5]^ during short-term follow-up of 6 months. Now, our experience of COR was extended to 18 clinical liver transplantations giving more precise insights into perfusion dynamics and biochemical parameters during rewarming of human livers from the cold to subnormothermic temperatures. Furthermore, long-term follow-up is now available for assessment of development of potential late complications and general safety up to 5 years after transplantation. Aim of the present study is to display our extended experience of now 18 clinical liver transplantations utilizing COR and demonstrate the long-term results of the recipients.
MATERIALS AND METHODS
=====================
Patients
--------
All data were collected prospectively. The first series of COR applications was performed between March 2014 and May 2014.^[@R5]^ Between June 2014 and June 2016 COR was applied in 12 additional allografts allocated to our center by EUROTRANSPLANT.
Indications for COR were given in cases of marginal allograft quality. All allografts were allocated by the rescue allocation mechanism of EUROTRANSPLANT after multiple refusals by several centers, indicating severely limited allograft quality. Underlying reasons were heterogeneous. The decision to accept these allografts for transplantation was made independently of the applied machine perfusion and thereby acquired hemodynamic or biochemical perfusion parameters.
Recipients with worse clinical condition than depicted by their laboratory model for end-stage liver disease scores (LabMELD) and without exceptional MELD scores have few chances to receive a liver allograft in a reasonable time. Such candidates were informed about the transplantation of marginal allografts, allocated by the rescue allocation mechanism by EUROTRANSPLANT, routinely during listing for LT and written informed consent was obtained. For the application of COR, every recipient was informed at the time of listing for transplantation and once again, when admitted for transplantation, that an individual treatment of the allograft was performed before transplantation and written informed consent was obtained.
The control group was created from all other consecutive liver transplantations performed during the same era (2014--2016) with similar characteristics regarding donor acceptance policies, recipient population, immunosuppression, surgeons, etc at our center. To achieve similar characteristics in both cohorts, we excluded recipients younger than 18 years of age, retransplantations, combined transplantations, living-related liver donations, split transplantations, and recipients with portal vein thrombosis.
Ethical Approval
----------------
The Ethics Committee of the University Hospital Essen approved the underlying study procedures (13-5409-BO).
Machine Perfusion (COR)
-----------------------
All donor allografts underwent the standard procurement procedures of cold storage at distant hospitals were preserved in either the University of Wisconsin or histidine tryptophan ketoglutarate (Custodiol) solution and then sent to our clinic. Upon arrival, allografts were revisited by an experienced transplant surgeon for suitability of transplantation. COR was performed in a standardized fashion during induction of anesthesia and hepatectomy. Detailed description is given elsewhere.^[@R6]^ The protocol of COR was designed with intention to avoid prolongation of overall cold ischemia time of the allograft. Back table preparation was performed as usual. Hepatic artery and portal vein were cannulated and separately perfused. Pulsatile perfusion pressure was used for hepatic artery, while continuous perfusion pressure was used for the portal vein. Pressures were set at 25 mm Hg (60 bpm) and 2--4 mm Hg, respectively. For machine perfusion, 2 L of Custodiol-N (Dr Köhler Chemie, Bensheim, Germany) was utilized and the solution actively oxygenated with 100% oxygen. Perfusion was started at 10°C and gradually increased to 12°C, 16°C, and 20°C after 30, 45, and 60 minutes, respectively. At the end of the COR procedure, immediately before implantation into the recipient the temperature of the allograft was reduced to approximately 16°C by flush with 1 L of cold (4°C) Custodiol solution. This approach presumably reduces warm ischemic damage during implantation. The core temperature was not measured in this series of COR applications. Preclinical data (not shown) demonstrated reduction of the core temperature of liver grafts to approximately 14°C to 16°C. Perfusion flows of hepatic artery and portal vein were measured during perfusion and perfusate samples collected. Laboratory investigations included amongst others oxygen pressure, pH, and lactate, as well as analysis of aminotransferases.
Surgery and Immunosuppression
-----------------------------
Procurements were performed in standard fashion as defined by EUROTRANSPLANT.^[@R7]^ Transplantation was performed with caval replacement and end-to-end-anastomoses of the hepatic artery, portal vein, and bile duct in all cases. Bypass techniques are not used at our center. The regimen of immunosuppression was standardized utilizing intravenous corticosteroids intraoperatively, followed by a standardized reduction scheme, calcineurin inhibitors (tacrolimus, trough level 6--8 ng/mL), and mycophenolate mofetil (0.5--1 g, twice daily). All patients were sent to intensive care unit (ICU) after transplantation and monitored closely. After discharge, all patients were followed by our outpatient clinic on a regular basis with ultrasound and fibroscan of the liver as routine procedures.
Fibroscan
---------
A representative area of liver parenchyma of the right lobe was identified by B-mode ultrasound. Hepatic elasticity was evaluated at this area by transient ultrasound elastography. Mean values of 10 measurements are reported.
Definition of Early Allograft Dysfunction
-----------------------------------------
The definition by Olthoff et al^[@R8]^ from 2010 was utilized: early allograft dysfunction (EAD) was diagnosed if bilirubin was 10 mg/dL or greater on postoperative day 7 and international normalized ratio was 1.6 or greater on postoperative day 7 and aspartate aminotransferase or alanine aminotransferase \>2000 IU/L within the first 7 days. Currently, EAD represents the best-validated and clinically relevant parameter for early liver allograft assessment after LT.
Definition of Ischemic Type Biliary Lesion
------------------------------------------
Ischemic type biliary lesions (ITBLs) were defined as intrahepatic biliary strictures and dilatations on a cholangiogram, in the absence of hepatic artery thrombosis. Diagnosis of ITBL was predominantly based on endoscopic retrograde cholangiogram (ERC) findings in the present study. ERC was performed as clinically indicated in cases of suspected biliary complications after exclusion of graft rejections. Routine magnetic resonance cholangiopancreatographies are not performed at our center.
Statistics
----------
Data were shown as the mean and SEM, or the median and range, as appropriate. Graft and patient survival were analyzed using the Kaplan-Meier analysis and compared using the log-rank test. A *P* \< 0.05 was considered statistically significant. We limited the comparison of the COR group with the control group to the most relevant endpoints (survival, EAD, ITBL, and length of stay) to avoid statistical issues due to multiple testing and underpowered comparisons. SPSS version 24 (IBM, Armonk, NY) was used for statistical analysis.
RESULTS
=======
Donors and Recipients
---------------------
Donor and recipient demographics are given in Table [1](#T1){ref-type="table"}. Indications for transplantation were given for alcoholic liver cirrhosis in 10 (55.6%), hepatitis B induced liver cirrhosis in 2 (11.1%), hepatitis C induced liver cirrhosis in 1 (5.5%), and other reasons in 4 (22.2%) cases. An HCC was diagnosed in 3 (16.7%) cases. Allograft steatosis was determined by prospectively collected biopsies. Macrovesicular steatosis was \<40% in all organs, concomitant to our local center acceptance policy. The median donor risk index (DRI) was 1.8 (1.03--2.5) in allografts undergoing machine perfusion. The median allograft weight was 1610 (1190--2570) grams. Reasons to select allografts for the machine perfusion protocol were as follows: extensive donor risk profiles in 5 cases with DRI \>2, extensive long cold ischemia time (\>12 h) in 6 cases, severe steatosis of the allograft diagnosed by biopsy in 5 cases, liver fibrosis in combination with poor macroscopic organ quality in 2 cases, and in 1 case additional brain malignancy (anaplastic astrocytoma) in combination with severe atherosclerosis of the abdominal arterial vessels leading to extended criteria donation.
######
Donor and recipient characteristics of COR treated patients
Machine Perfusion
-----------------
All livers were perfused in accordance to the COR protocol without incidences and transplanted successfully. Arterial perfusion flows and portal vein perfusion flows are depicted in Figure [1](#F1){ref-type="fig"}. The temperature slope of rewarming was very similar in all allografts, independent from the allograft weight. After 30 minutes of cold perfusion mean arterial perfusion flow was 109 ± 15.1 mL/min rising to 169 ± 20.4 mL/min at 120 minutes of perfusion and 20°C. The portal vein flow increased from 118 ± 24 to 139 ± 45 mL/min at those time points. Lactate concentrations were 2.5 (0.9--4.9) mmol/L after 30 minutes of perfusion and 3.25 (1.1--6.7) mmol/L at the end of perfusion. Concentration of aminotransferase aspartate aminotransferase was 998 (131--3935) U/L 30 minutes after induction of perfusion and increased to 1645 (244--5328) U/L after 120 minutes of perfusion and elevation of the temperature to 20°C. Bile production was minimal during 120 minutes of perfusion and was, therefore, not quantified in detail.
{#F1}
Postoperative Outcomes
----------------------
An overview of outcome data after application of COR before LT is given in Table [2](#T2){ref-type="table"}. Detailed postoperative laboratory courses are given in Figures [2](#F2){ref-type="fig"} and 3. EAD was observed in 2 (11.1%) cases after COR and based on elevations of aminotransferases \>2000 U/L within the first 7 days after LT. Primary nonfunction (PNF) was not observed. No patient required retransplantation. Length of the first stay (if readmission to ICU only the first stay was counted) in the ICU was median 5 days and hospital stay was median 21 days. An ITBL was observed in 1 patient (5.5%) after transplantation with preceding COR. The donor liver of this recipient was afflicted with an extensive cold ischemia time of \>14 hours. This patient was treated by multiple ERCs and stenting of the biliary system in the further follow-up; however, he died due to progressive disease and recirrhosis. All other patients are well and alive. The according 1-, 3-, and 5-year patient survival rates are 100%, 100%, and 93.8%, respectively.
######
Clinical outcome after COR
{#F2}
Fibroscan
---------
All patients underwent routine fibroscan investigations at routine visits at our outpatient clinic. Liver elasticity was median 4.65 (2.8--14.8) kPa at the last follow-up in patients without ITBL, indicating normal healthy liver parenchyma.
Comparison With Control Cohort
------------------------------
For further classification of the presented results of COR before LT, we generated a control cohort based on similar clinical characteristics. Comparison with the COR group showed a significantly higher LabMELD score in the control group (Table [3](#T3){ref-type="table"}) and, on the other hand, a significantly longer cold ischemic time (CIT) in the COR group. No baseline characteristics showed relevant differences between groups.
######
Comparison of COR treated patients with controls
We limited the comparison of the COR group with the control group to the most relevant endpoints (EAD, ITBL, length of stay, and survival) to avoid statistical issues due to multiple testing and underpowered comparisons. Rate of EAD after LT was numerically lower in the COR group compared to the control group without statistical significance. Rate of PNF and retransplantation did not qualify for statistical comparisons, due to low number of events. Application of hemodialysis after LT was similar in both groups as well as length of ICU stay and hospital stay. ITBLs were observed in similar frequency in the COR group and control group. Comparison of graft survival demonstrated superior outcomes after COR without statistical significance (*P* = 0.12): The 1-, 3-, and 5-year patient survival rates for the COR and control groups are depicted in detail in Figure [4](#F4){ref-type="fig"}.
{#F3}
{#F4}
DISCUSSION
==========
This study displays our extended experience of application of COR as end-ischemic machine perfusion modality in 18 recipients before clinical liver transplantation. The presented results demonstrate first of all safety of this new technique in clinical liver transplantation for short- and long-term follow-up.
All donor organs which were selected for the COR procedure had extensive risk profiles, resembled by a median DRI of 1.8. Organs were allocated by the rescue allocation mechanism by EUROTRANSPLANT indicating that these were organs that nobody wanted. However, some organs had a low DRI but were afflicted with other risk factors like distinct steatosis, which is not reflected by the DRI. Perfusion characteristics were in line with our preclinical and clinical experience: portal and arterial perfusion flows increased in accordance to elevation of perfusion temperature. Lactate concentrations rose moderately during perfusion. Aminotransferase release was highest in the first 30 minutes, thereafter increasing moderately in individual ranges for each allograft. Release of aminotransferases has been shown to correlate well with posttransplant hepatocellular injury^[@R9]^ and might be a useful parameter to predict suitability of the allograft before transplantation. However, investigations in a larger number of patients are needed to prove this observation.
After transplantation, we detected only mild to moderate hepatocellular injury after COR treatment, displayed by aminotransferase release as usually seen in optimal allografts (Figure [2](#F2){ref-type="fig"}). Accordingly, liver function stabilized quickly represented by laboratory values of bilirubin and international normalized ratio, which nearly normalized in the first 7 days after transplantation in all recipients (Figure [3](#F3){ref-type="fig"}). PNF was not observed. These results clearly support our first experience of COR after cold storage before LT, which was published in 2016.^[@R5]^ The clinically important end point of EAD^[@R10]^ was observed in 2 out of 18 patients (11.1%) only after COR. Accordingly, the outcome in the first months after LT was excellent with survival rates after 1, 3, and 6 months of 100% (Figure [4](#F4){ref-type="fig"}).
During the further long-term follow-up 1 patient was diagnosed with ITBL and treated repeatedly by ERC, dilatation, and stent implantation.^[@R11]^ This patient died due to biliary complications and recirrhosis nearly 3 years after LT. All other patients have no evidence of ITBL and are well and alive after a median follow-up of 1546 (749--1933) days, demonstrating excellent long-term outcomes in spite of the elevated risk afflicted with the utilized allografts. Such outcome data are far superior to the usual results in the EUROTRANSPLANT area with survival rates of ≈80% and 65% after 1 and 5 years, respectively.^[@R12]^
With the knowledge that survival after LT is multifactorial^[@R13],[@R14]^ it can be questioned whether recipient selection impacts the presented results: Recipient characteristics demonstrate that fair candidates were selected as recipients for this series of COR applications in terms of the LabMELD scores.
For further classification of our presented COR results, we created a control cohort with similar clinically relevant parameters influencing postoperative results: all control patients were adults undergoing their first liver transplantation and we excluded combined transplantations, living-related liver donations, split transplantations, and recipients with portal vein thrombosis. We want to emphasize that the presented study procedures were not designed as a controlled study and no power analysis or sample size calculations were performed. In this context, we compared only the most important clinical endpoints after transplantation with intention to abstain from statistical issues of multiple testing. Comparisons of both groups showed a higher LabMELD score in the control group, which most likely demonstrates a selection bias of the COR patients. As the MELD score predicts posttransplant survival in some studies the reported results should be construed against this background and presumably limits the interpretation of the results in favor of COR. We cannot rule out that the difference in survival observed in this study is (at least in part) due to the different MELD scores and not only associated with COR. On the other hand, the CIT was significantly longer in the COR group adding relevant risks to the profiles of these organs.^[@R14],[@R15]^ All other comparisons of the COR group with the control group demonstrated similarity in recipient, donor, and procedure characteristics (Table [3](#T3){ref-type="table"}) as expected. Investigated outcome parameters did not show any statistical difference between groups (Table [3](#T3){ref-type="table"}). The graft survival demonstrated clinically relevant superior results after COR (Figure [4](#F4){ref-type="fig"}), suggesting a protective effect of COR, however, not reaching statistical significance. Although nonsignificant, the numerical difference of EAD between groups might be a major reason for this result, especially as graft survival decreased in the early period after transplantation in the control group. Early graft failure is associated with suboptimal liver function after transplantation, leading frequently to infectious complications with deteriorating liver function and graft failure.^[@R10],[@R16]^ Definite superiority of COR in comparison to other preservation strategies needs to be further investigated in a randomized controlled trial, which is already initiated and recruiting at our center (ISRCTN 94691167).
The rate of ITBL was similar in both groups with ≈5%. This complication is feared after LT and presumably based on suboptimal microperfusion of biliary capillaries, therefore, injury of peribiliary glands and vascular plexus resulting in suboptimal preservation of ischemia sensitive biliary epithelial cells.^[@R17],[@R18]^ Unfortunately, destruction of the biliary tree results leading to formation of biliary cast, cholestasis, cholangitis, and finally recirrhosis.^[@R19]^ It is of major interest to reduce this complication and perfusion modalities are thought to improve the capillary flow and energy supply of the tissue, theoretically reducing the risk for ITBL. However, the results of the present study do show a case of ITBL in the COR group and, therefore, do not support this in the given very small sample size and, therefore, underpowered analysis. Other perfusion modalities like normothermic machine perfusion demonstrated a rate 7.5% of nonanastomotic strictures after machine perfusion compared to 5.4% after static cold storage in line with the incidence of the study at hand.^[@R20]^ Future studies with sufficient numbers of patients to investigate this question are of high interest, however, difficult to accomplish due to the high number of patients needed for a sufficiently powered analysis. So far, data from clinical applications are the only source to give more insights on this very important topic.
The concept and physiology of COR have been investigated thoroughly in preclinical studies.^[@R21]^ It was demonstrated that key for integrity of the parenchyma and function of the allograft is the slow elevation of the temperature and not only the simple increased mean temperature during perfusion.^[@R21]^ Interestingly, the flush of the liver at the end of COR with cold preservation solution reduces the temperature to 14°C to 16°C only. Therefore, the washout of the machine perfusate has only little influence on the core temperature of the graft. A recent case report demonstrated that it might be feasible to waive this maneuver before implantation in clinical practice,^[@R22]^ but this remains a matter of future investigations. In our understanding, the abrupt rewarming injury, which might be associated with such flush out maneuver, is most relevant after longer periods of hypothermia but rather undetected when starting from temperatures around 16°C.^[@R23]^ So COR after cold storage restores energy homeostasis of the tissue at cold to subnormothermic conditions and optimally reconditions the allograft for the implantation warm ischemia, warm reperfusion with blood and the concomitant reperfusion injury. Experimental data suggest that this results in less oxidative injury, lower vascular resistance in the portal vein, and significant reduction of inflammation on reperfusion.^[@R6]^
The evolution of different perfusion modalities and adoption from laboratory to the clinic in the recent years, namely hypothermic machine perfusion,^[@R24]^ hypothermic oxygenated machine perfusion,^[@R25]^ and normothermic machine perfusion^[@R20]^ resembles the most relevant progress in the field of organ preservation for decades. While all have demonstrated feasible and promising results in clinical application so far it remains unclear which perfusion modality is superior to others or whether different strategies are needed for different allografts in dependence of risk profiles (eg, long CIT, steatosis, DCD, etc). Future studies will hopefully gain insight of these exciting questions.
Limitations of the present study include the monocentric approach and nonrandomized design. Furthermore, selection of allografts for the COR protocol was based primarily on the rescue allocation mechanism of EUROTRANSPLANT, resembling a heterogeneous group of organs of limited quality and adoption of the results to a certain risk group of organs is not suitable so far. Moreover, the control group differs significantly in terms of MELD score, potentially influencing the patient survival after transplantation in this group. As far as this study is based on a nonrandomized design, the impact of the intervention should be interpreted with caution.
In conclusion, the present study demonstrates excellent clinical long-term outcomes after application of COR before LT in the largest clinical data set available so far. Comparison with a similar control cohort shows superiority of graft survival. A randomized controlled clinical trial has been initiated and is recruiting to give further evidence of this promising method to improve organ preservation.
Published online 13 March, 2020.
The authors declare no funding or conflicts of interest.
D.P.H. participated in the research design, writing of the paper, performance of research, and data analysis. T.B., P.M., C.v.H., and J.W.T. participated in the performance of research. A.P. participated in the research design and writing of the paper. T.M. participated in the research design.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION
============
Idiopathic ventricular tachycardia (VT), which develops without any cardiac abnormalities, can be divided into four types: outflow tract VT, verapamil sensitive fascicular VT, focal Purkinje VT and annular VT. The main mechanism of idiopathic left VT is reentry that passes through anterior or posterior fascicle.[@B1],[@B2] This fascicular VT commonly shows left axis deviation of the right bundle branch block (RBBB) of more than 90%. On the other hand, focal Purkinje VT, which is initiated by automaticity, is very infrequent. Purkinje VT is characterized by a lack of response to verapamil and by its inability to being induced or terminated by electrical programmed stimulation. However, the exact clinical aspect or etiology of this VT has not yet been described. In this report, we describe a case in which catheter ablation was used to successfully treat a focal Purkinje VT combined with tachycardia-induced cardiomyopathy.
CASE REPORT
===========
A 36-year-old man with no specific past illnesses was referred for evaluation of his tachycardia and dyspnea for a week. After admission, physical examination revealed grade 2/6 systolic murmur at the left sternal borders and decreased breathing sounds with both basal rales. The initial electrocardiogram (ECG) revealed wide QRS tachycardia ([Fig. 1](#F1){ref-type="fig"}), and the chest X-rays showed marked cardiomegaly with pulmonary congestion. A routine blood profile, including myocardial markers and inflammatory markers, did not reveal any specific abnormalities. The wide QRS tachycardia observed in the ECG showed RBBB, left axis deviation and atrioventricular dissociation. Echocardiography showed 25% of left ventricular ejection fraction and enlargement of the left ventricle and left atrium without a regional wall motion abnormality. Intra-venous and oral verapamil was administered under the suspicion of tachycardia-mediated cardiomyopathy resulting from idiopathic posterior fascicular VT. However, continuous telemetry monitoring demonstrated incessant VT without any response to the treatment. Therefore, electrophysiological studies and radiofrequency catheter ablation (RF-CA) were carried out.
During the electrophysiological recordings, intermittent sinus capture beats were observed, nevertheless, VT continued, and it could not be terminated with 12 mg of adenosine or right ventricle overdrive pacing (pacing cycle length 260-200 ms). Ablation was started from the distal area, where the earliest Purkinje potential was found ([Fig. 2](#F2){ref-type="fig"}). Entrainment at ablation site did not meet the classical criteria, and pace-mapping at ablation site was similar but not exactly matched with clinical VT. It might be due to fusion of pacing beats with incessant VT. If the VT failed to stop within 20 seconds of RF-CA application, the ablation catheter was moved more proximally, and the RF-CA application was restarted. VT was terminated within about four seconds of RF-CA application at the LV mid septal area during 4^th^ RF energy application. Afterwards, a few ventricular beats were observed for 1-2 seconds before disappearing completely. Two more RF-CA applications were delivered at the same location. After the final RF-CA application, VT was not induced either by the programmed stimuli or an isoproterenol infusion of up to 20 µg/min.
After the procedure, there was no left bundle branch block (LBBB) or intraventricular conduction delay on the ECG. Telemetry monitoring revealed that VT did not return, and that normal sinus rhythm was maintained without a single premature ventricular contraction. The patient\'s symptoms were improved, and he was discharged two days after the procedure. Follow up echocardiography performed five months after catheter ablation showed 49% of left ventricular ejection fraction and shrinked LV. Chest X-rays performed in an outpatient department six months after discharge showed a decreased cardiac silhouette ([Fig. 3](#F3){ref-type="fig"}) without symptoms. Follow-up observations are ongoing, and neither VT nor any cardiac symptoms have recurred.
DISCUSSION
==========
Because this young patient had pulmonary congestion and cardiomegaly at the time of his visit, myocarditis-induced heart failure with VT was considered also as a possible diagnosis. Ventricular arrhythmias complicating fulminant myocarditis can be initiated and maintained by activity originating from the Purkinje system, and can be suppressed by catheter ablation.[@B3]-[@B6] However, a differential diagnosis was made based on the following facts.[@B2]-[@B5] First, the initial ECG showed a relatively narrow QRS duration (126 ms), which is not consistent with myocardial inflammation, in which it is usually prolonged to more than 140 ms. Second, the patient\'s detailed medical history did not indicate any upper respiratory tract infections or flu-like symptoms before the visit to local clinic. Finally, the patient tested negative for all myocardial inflammatory markers that were evaluated in our hospital. In addition, the typical morphology which we observed was compatible with posterior fascicular VT. Therefore, we were able to diagnose the patient with tachycardia-induced cardiomyopathy due to incessant VT originating near or from the Purkinje system. Usually, the known phenomena underlying posterior fascicular VT is macro-reentry sensitive to verapamil. However, this case did not show any response to verapamil, and it did not terminate with programmed electrical stimulation or adenosine. Purkinje fibers can be bystander exit sites, and Purkinje potential is not hallmark of focal Purkinje VT. However, a small number of energy application required to eliminate the VT suggested that the origin of VT was Purkinje fiber.[@B3]-[@B6] In view of these observations, the patient was diagnosed as a focal Purkinje VT, due to automaticity rather than triggered activity. Focal Purkinje VT is such a rare disease, so that its clinical and electrophysiological characteristics are not well known. It can appear in two forms: RBBB or LBBB.[@B2],[@B3] It is also known to respond to propranolol, a beta blocker. Additionally, it can be induced possibly by exercise or catecholamines, however, it cannot be induced or terminated by programmed electrical stimulation. Given the fact that Purkinje fibers play a key role in initiating and maintaining the VT and they are located in the subendocardium, RF-CA could be a good treatment for this type of arrhythmia.
This work was supported in part by Yonsei University Research Fund of 6-2008-0009 (JB Kim).
The authors have no financial conflicts of interest.
{#F1}
{#F2}
{#F3}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
*Mycobacterium tuberculosis* (*Mtb*) is responsible for the greatest number of deaths by a bacterial pathogen worldwide [@pone.0058378-World1]. *Mtb* can produce progressive disease or latent infection [@pone.0058378-Parrish1]; indeed, in high endemic areas, infection first occurs in childhood and is controlled in most cases. Only 10% of these primary infections lead to progressive disease [@pone.0058378-Parrish1], [@pone.0058378-Tufariello1]; however, some bacilli remain in tissues in a non-replicating dormant or slowly replicating stage for the rest of the individual's life. This latent tuberculosis (LTB) is clinically asymptomatic, and in countries with low or moderate endemicity, most active tuberculosis (TB) cases arise as a result of the reactivation of latent bacilli [@pone.0058378-Parrish1], [@pone.0058378-Tufariello1]. It is estimated that one third of the world's population carries latent *Mtb*, and millions of TB reactivation cases are predicted to occur in the coming years [@pone.0058378-Kaufmann1].
It is widely accepted that cell-mediated immunity, primarily by Th-1 cytokines and TNF-α, controls LTB and that dormant bacilli persist in granulomas under hostile conditions [@pone.0058378-Fritz1]--[@pone.0058378-Wayne1]. Experimental models of LTB have been performed *in vitro* to simulate the conditions faced by latent bacilli in granulomas, such as low-oxygen tension, starvation or acid pH, and *in vivo* using animal models to closely simulate the immunological response related to LTB in mammal tissues. The hypoxic "Wayne" model has been widely used as an *in vitro* model in which oxygen is gradually withdrawn, thereby inducing mycobacteria to enter into a non-replicative persistence (NRP) state [@pone.0058378-Wayne2]. *In vivo*, it is possible to induce progressive disease or latent infection by the intratracheal inoculation of a high or low infecting dose of *Mtb* in BALB/c or C57bl/DBA mice, respectively [@pone.0058378-Arriaga1], [@pone.0058378-HernndezPando1]. This LTB model is reproducible and it is characterized by granuloma formation, high expression of TNFα, iNOS, IL-2 and IFNγ without tissue damage (pneumonia) displaying low and stable lung bacillary loads [@pone.0058378-Arriaga1]. In contrast, progressive TB is produced by the intratracheal administration of high infecting dose that ensures bacilli proliferation, formation of granulomas with mild interstitial and perivascular inflammation raising maximal Th-1 protective response at day 21 of infection, followed by progressive bacilli burdens, in coexistence with tissue damage and emergence of Th-2 cells [@pone.0058378-Arriaga1], [@pone.0058378-HernndezPando1].
An important characteristic of latent bacilli is their gradual loss of acid-fastness associated with variations in the cell wall microstructure, suggesting alterations in lipid composition [@pone.0058378-Seiler1]. In addition, hypoxia induces the transcription of genes involved in the biosynthesis and production of some cell-wall glycolipids [@pone.0058378-Starck1]--[@pone.0058378-Yu1] and mycolic acids [@pone.0058378-Starck1]. Trehalose-based glycolipids are significant constituents of the mycobacterial cell wall, with diacyltrehalose/polyacyltrehalose (DAT/PAT) and sulfolipid (SL) being important members of this group; the mechanisms of SL and DAT/PAT biosynthesis have been recently proposed [@pone.0058378-Dubey1]--[@pone.0058378-Domenech1]. Thus, the polyketide synthases (Pks) 2, 3 and 4 synthesise acyl chains that esterify hydroxyl groups of trehalose by the action of the PapA1, PapA2 and PapA3 acyltransferases ([Figure 1](#pone-0058378-g001){ref-type="fig"}). Subsequently, the MmpL8 and MmpL10 membrane-associated transporters export these synthesised lipids or their precursors to the external side of the cell wall.
{#pone-0058378-g001}
The *Mtb* two-component system PhoP-PhoR positively regulates the synthesis of SL, DAT and PAT during bacilli phagocytosis; therefore, disruption of *phoP* in *Mtb* yields mutants lacking trehalose-based glycolipids, which is also observed in the attenuated *Mtb* strain, H37Ra [@pone.0058378-GonzaloAsensio1]. Regarding their function in the host, SL inhibits mitochondrial oxidative phosphorylation and blocks phagosome-lysosome fusion in macrophages; SL also modulates the oxidative response and the secretion of anti-inflammatory cytokines (IL-10 and IL-13) by human monocytes and neutrophils [@pone.0058378-Brennan1]--[@pone.0058378-Gilleron1], and accumulation of SL precursors can stimulate human CD1b-restricted T cells [@pone.0058378-Gilleron1].
Several studies have shown that *Mtb* H37Rv knock-out mutants of *pks2* are defective in SL biosynthesis; however, these strains do not show a significant decrease in persistence or pathogenicity in mice or macrophages [@pone.0058378-Jackson1]. *Mtb mmpL8* knock-out mutants produce diacylated forms of SL and are less virulent in mice. *In vitro* studies have shown that mycolipenic and mycoseric acids, the acyl substituents in DAT and PAT, inhibit leukocyte migration, but mutants that do not synthesise mycolipenates and mycosanoates do not show virulence alteration in macrophages or mice [@pone.0058378-Brennan1], [@pone.0058378-Jackson1]. The actual role of trehalose-based glycolipids in LTB is completely unknown. Thus, the aim of the present study was to evaluate the transcriptional behaviour of genes involved in the SL and DAT/PAT biosynthesis in *in vitro* and *in vivo* experimental models of LTB. We observed that the genes involved in SL and PAT biosynthesis of *Mtb* are differentially expressed in experimental progressive and LTB infection.
Materials and Methods {#s2}
=====================
Ethics Statement {#s2a}
----------------
All the animal work was done according to the guidelines and approval of the Ethical Committee for Experimentation in Animals of the National Institute of Medical Sciences and Nutrition in Mexico (CINVA), permit number: 224. All surgery was performed under sevofluorane anaesthesia, and all efforts were made to minimize suffering.
Mycobacterial Strains and Growth Conditions {#s2b}
-------------------------------------------
*Mtb* H37Rv (ATCC 27294, Rockville, MD, USA) was used in this study. Mycobacteria were cultured at 37°C in Dubos-ADC (Difco) until the exponential phase of growth was reached (standard conditions). The cultures were then subjected to non-replicative persistence 1 (NRP1) and non-replicative persistence 2 (NRP2) or anaerobiosis stages as described by Wayne and Hayes [@pone.0058378-Wayne2]. A parallel culture supplemented with methylene-blue (1.5 µg mL^−1^) was used as an indicator of oxygen depletion. Mycobacteria grown in standard and hypoxic (NRP1 and NRP2) conditions were used for RNA isolation and lipid analysis. All determinations were performed at least in duplicate.
Lipid Extraction and Analysis by TLC {#s2c}
------------------------------------
Glycolipids were extracted as previously described [@pone.0058378-Muoz1] from oxygenated mycobacteria harvested in exponential phase of growth and NRP2 stage. The SL and DAT content was analysed on 20×20 Silica Gel 60 thin-layer chromatography (TLC) plates (Merck, Germany) by two-dimensional TLC [@pone.0058378-Soto1]. Mycolic acids were analysed as described previously [@pone.0058378-Luquin1]. Carbohydrate-containing compounds were visualised by spraying TLC plates with 1% anthrone (Sigma, USA) in H~2~SO~4~, followed by heating at 120°C. Methyl mycolates were visualised using 10% molybdophosphoric acid (Merck, Germany) in ethyl alcohol (w/v) and heating at 120°C. *Mtb* H37Rv (ATCC 27294) SL-I, DAT and CF standards were developed and visualized in parallel experiments.
Neutral-red Test {#s2d}
----------------
*Mtb* cells were placed in screw-cap tubes containing 50% (v/v) aqueous methanol and incubated for 1 h at 37°C. The fluid was then removed, 0.002% neutral red in Tris-HCl buffer at pH 9.8 was added, and the tubes were kept at room temperature for 24 h [@pone.0058378-Soto2].
Experimental Models of Latent Mycobacterial Infection and Progressive Pulmonary Tuberculosis in Mice {#s2e}
----------------------------------------------------------------------------------------------------
The models of latent and progressive TB in mice have been described elsewhere [@pone.0058378-Arriaga1], [@pone.0058378-HernndezPando1]. The *Mtb* strain H37Rv was grown until an OD~600~ of 0.4--0.8 was reached. Bacilli were harvested, adjusted to 1×10^3^ bacteria/100 µL in phosphate buffered saline (PBS), aliquoted and maintained at −70°C until use. The viability of the bacteria was checked before infection. Induction of chronic infection with a long-lasting stable infection similar to LTB was performed with 6--8-week-old female B6D2F1 (C57BL/6J×DBA2/J) mice (Jackson Laboratories Bar Harbor, ME, USA). Mice were anesthetised with sevofluorane and inoculated intratracheally with 1×10^3^ viable bacilli in 100 µL of PBS. In addition, a group of mice with stable latent TB after 7 months of infection were supplemented with corticosterone (3 mg mL^−1^) dissolved in their drinking water to induce reactivation. Groups of three mice were killed by exsanguination at 1, 3 and 5 months after intratracheal infection and 20 days after corticosterone supplementation.
To induce progressive pulmonary TB, 6--8-week-old male BALB/c mice were anesthetised with sevofluorane and inoculated intratracheally with 2.5×10^5^ bacilli in 100 µL of PBS (treatment of bacteria was similar to the LTB model). Groups of three mice were killed by exsanguination at 1, 28 and 120 days after intratracheal infection. All the animal work was done according to the guidelines and approval of the Ethical Committee for Experimentation in Animals of the National Institute of Medical Sciences and Nutrition in Mexico (CINVA), permit number: 224. All surgery was performed under sevofluorane anaesthesia, and all efforts were made to minimize suffering.
RNA Isolation and cDNA Synthesis {#s2f}
--------------------------------
RNA from *Mtb* was extracted using TRIzol reagent (Invitrogen, USA) as described elsewhere [@pone.0058378-Shi1], [@pone.0058378-GonzalezyMerchand1]. For *in vitro* analysis, total RNA from standard *Mtb* cultures was isolated in the exponential and stationary phases of growth; for hypoxic conditions, RNA was isolated from cultures in NRP1 (after 5 days of exposure to hypoxia) and NRP2 (after 11 days of exposure to hypoxia) stages of Wayne's model of dormancy. Total RNA was dissolved in DEPC- treated H~2~O and stored at −70°C. A previously described technique was used to determine mycobacterial gene expression by quantitative real-time PCR (qRT-PCR) after the isolation of mycobacterial RNA from tissues [@pone.0058378-Shi1]. For each time point of the *in vivo* infection TB models, lungs from mice were perfused with 1 mL of lysis buffer (6 M guanidinium chloride, 0.1% Tween 80, 10 mM EDTA and 1 mM mercaptoethanol) and frozen by liquid nitrogen immersion. These tissues were kept at −70°C until use. Lungs were disrupted in o-ring tubes containing 2 ml of lysis buffer and zirconium-silica beads (710--1180 and 150--212 µm, Sigma) in Fast-Prep equipment (Thermo, Germany). Total RNA from these homogenates was isolated with the same methodology used for mycobacteria in culture. cDNA from all samples was prepared using 2 µg of RNA, random hexamers (0.5 µg/µl) and Super Script® reverse transcriptase (20 U µL^−1^, Invitrogen).
Quantitative Real-Time PCR (qRT-PCR) {#s2g}
------------------------------------
The genes involved in SL biosynthesis (*pks2*, *papA1*, *papA2* and *mmpL8*) and DAT/PAT biosynthesis (*pks3*, *pks4*, *papA3*, *mmpL10*), together with the *phoP* and *rfpB* genes, were quantified by qRT-PCR using specific primers ([Table 1](#pone-0058378-t001){ref-type="table"}). Total mRNA transcription was normalised to the mean value of 16S rRNA (*rrs*) gene expression [@pone.0058378-BadilloLpez1]. Quantitative differences in transcription among growth phases and periods of *in vivo* infection were analysed with logarithmic graphs of each gene (reported as copies of the gene/µg RNA). Fluorescence was quantified using LC FastStart DNA Master SYBR Green I (Roche) and the LightCycler® 1.5 system (Roche). To validate qRT-PCR assays and determine their efficiency, serial dilutions of genomic DNA of *Mtb* H37Rv were tested. Quantification was performed four times for each condition, and results were analysed using a calibration curve for which the regression value was close to 1 and the efficiency was close to 2.
10.1371/journal.pone.0058378.t001
###### Primers for qRT-PCR used in this study.
{#pone-0058378-t001-1}
Primer Sequence (5′--3′) Reference
----------- ----------------------- ------------------------------
*pks2*F CGGTGACCCCATTGAATAT This study
*pks2*R CACCATGTTTCAGAGCGAGA This study
*papA1*F GATGCTAATGGACGAGCAG This study
*papA2*R CGTCACTGTGGTCTGATGCT This study
*mmpL8*F GGCGGGGTTCTATATTCC This study
*mmpL8*R GGATTCAGGTCGGTTTGTAT This study
*pks3*F GATCAGCTGGCTGAGATTG This study
*pks3*R GCCGCTGGTTCTGATAC This study
*pks4*F TATGGGGTACGTCCTGGTG This study
*papA3*F GCGATGTTCCGGTCAGTT This study
*papA3*R GTACGCCGGATTATTTGTCC This study
*mmpL10*F GGCTTGACTCTGTTGTCGC This study
*mmpL10*R GATTTCAGCGAGCGGACTA This study
*phoP*F TCTCGACCACGTTTGGCGCT This study
*phoP*R CCCGCAGTACGTAGCCCACC This study
*rpfB*F CCGCAATCGGATCAAGAA [@pone.0058378-BadilloLpez1]
*rpfB*R CGACCTCCCGGCTCAT [@pone.0058378-BadilloLpez1]
16SrRNAF ATGACGGCCTTCGGGTTGTAA [@pone.0058378-BadilloLpez1]
16SrRNAR CGGCTGCTGGCACGTAGTTG [@pone.0058378-BadilloLpez1]
Statistics {#s2h}
----------
For each experiment, differences among experimental data were compared using Two-Way ANOVA, Tukey's multiple-comparison procedure. Data were accepted as significantly different if *P*\<0.05.
Results {#s3}
=======
Hypoxia Induces Change in the Trehalose-based Glycolipid Profile of *Mycobacterium tuberculosis* H37Rv {#s3a}
------------------------------------------------------------------------------------------------------
Cultures of *Mtb* obtained under hypoxic conditions produced more dispersed mycobacteria, indicating a possible alteration of the lipid content on the bacterial surface. Non-covalently attached lipids were extracted from *Mtb* grown in oxygen-supplied and NRP2 conditions. As an initial approach, analysis of crude lipid extracts by TLC revealed similar profiles of phospholipids and phosphatidylinositol mannoside (PIM) under both conditions. Accumulation of triacylglicerol (TAG) in NRP2 was also observed (data not shown). The glycolipid TLC profile of *Mtb* H37Rv in NRP1 was similar to that observed in bacilli grown under standard conditions (data not shown). According to the purified glycolipid standards used in the 2D-TLC analysis, trehalose-based glycolipids in NRP2 showed deficient production of SL, an increased DAT content, and the accumulation of polar glycolipids with chromatographic behaviour similar to PAT ([Figure 2](#pone-0058378-g002){ref-type="fig"}).
{#pone-0058378-g002}
With regard to the mycolic acid content, oxygenated and hypoxic (NRP2) cells showed similar pattern of mycolates; the characteristic α (I), methoxy (III) and keto-mycolates (IV) for *Mtb* were detected under both growth conditions (data not shown). Finally, *M. tuberculosis* in NRP2 did not become red when cells were subjected to neutral red staining ([Figure 2F](#pone-0058378-g002){ref-type="fig"}).
*Mycobacterium tuberculosis* Displays a Differential Transcription of *mmpL* and *pks* Genes in the NRP2 Stage {#s3b}
--------------------------------------------------------------------------------------------------------------
RNA was isolated from two replicas of standard (oxygenated) *Mtb* H37Rv cultures in the exponential and stationary phases of growth and hypoxic (NRP1 and NRP2) conditions to quantify the transcription pattern of the selected genes by qRT-PCR. These mycobacterial genes related to glycolipids production were chosen according the gene clusters previously proposed for the SL-I and DAT/PAT biosynthetic pathways (See [Figure 1](#pone-0058378-g001){ref-type="fig"}) [@pone.0058378-Hatzios1]. At the present time, the biosynthetic pathway of DAT is not well-known.
When compared with bacteria in the exponential phase, oxygenated bacilli grown until the stationary phase exhibited increased transcription of *papA1* and *papA2,* both involved in the SL biosynthesis, and decreased transcription of *mmpL10* that codify for the transporter in the DAT/PAT biosynthesis ([Figure S1](#pone.0058378.s001){ref-type="supplementary-material"}). Under hypoxic conditions, the highest differences in transcription were observed in NRP2 in comparison to oxygenated bacilli in the exponential phase. With respect to the SL biosynthesis, the relative quantification showed that the transcription of *pks2*, *papA1* and *papA2* genes was 19-, 17- and 8-fold higher in NRP2 respectively. Conversely, the transcription rate of the *mmpL8* gene was 3-fold lower in NRP1 and extremely low (more than 8×10^3^-fold) in NRP2 ([Figure 3](#pone-0058378-g003){ref-type="fig"}). With regard to the DAT/PAT biosynthesis, the transcription of *pks3/pks4* was 10- to 11-fold higher, respectively, in NRP2. The transcription rate of the *mmpL10* gene was 5- and 2-fold lower in NRP1 and NRP2, respectively. Finally, the transcription of the *phoP* regulator was 2- and 10-fold higher in NRP1 and NRP2, respectively ([Figure 3](#pone-0058378-g003){ref-type="fig"}).
{#pone-0058378-g003}
Differential Transcription of Genes Involved in SL and DAT/PAT Biosynthesis in Murine Models of Progressive and Latent Tuberculosis Infection {#s3c}
---------------------------------------------------------------------------------------------------------------------------------------------
In the chronic infection model that is similar to LTB, bacilli loads were constant (nearly 1×10^5^, data not shown). For relative quantification, the transcription level of the genes was compared with the transcription level on day 28 of the progressive disease. The transcription of some genes involved in the biosynthesis of SL and PAT (see [Figure 1](#pone-0058378-g001){ref-type="fig"}) increased during the course of experimental LTB, particularly *mmpL8* and *mmpL10*; however, the *pks3/pks4* genes exhibited slightly diminished transcription at month 5 in the LTB model ([Figure 4](#pone-0058378-g004){ref-type="fig"}). After LTB reactivation, all the studied genes exhibited significantly increased transcription. Reactivation was confirmed by the increased activity of the control gene *rfpB* ([Figure 4](#pone-0058378-g004){ref-type="fig"}), which was associated with high bacilli loads in the lungs. The transcription of *phoP* was always high and constant during the course of latent infection, reactivation and progressive disease ([Figure 4](#pone-0058378-g004){ref-type="fig"}).
{#pone-0058378-g004}
Regarding the progressive TB model, lung bacilli load determinations by CFU quantification showed that the initial bacilli load (2.5×10^5^) increased to 9.5×10^6^ at day 28 and 1.4×10^7^ at day 120 (data not shown). The transcription of most the selected genes strongly increased during the course of progressive infection ([Figure 5](#pone-0058378-g005){ref-type="fig"}). On day 1, the relative transcription of *mmpL8*, *papA3, pks3/pks4* and *phoP* was considerably less than for the other genes (2- to 60-fold lower); however, transcription of these genes, except *mmpL10*, increased considerably by day 120 ([Figure 5](#pone-0058378-g005){ref-type="fig"}). The values of relative quantification were based on absolute normalised quantification values showed in figures S2 and S3.
{#pone-0058378-g005}
Discussion {#s4}
==========
Because trehalose-based glycolipids such as SL and PAT are only present in virulent *Mtb* strains and therefore potentially play a role in virulence, recent studies have focused on establishing the actual function of this kind of glycolipid in TB pathogenesis [@pone.0058378-Dubey1]--[@pone.0058378-Domenech1]; however, their role in LTB remains unknown. A high bacterial load is necessary for a complete mycobacterial lipid constitution analysis, which is impossible to obtain from infected tissues. We therefore used the *in vitro* Wayne model to analyse the SL and DAT profile of the hypoxic *Mtb* cell wall. Although *in vitro* models do not completely reproduce the conditions of LTB, our results using the Wayne model showed previously unidentified alterations to the trehalose-based glycolipid composition of hypoxic *Mtb*. The murine model of chronic infection similar to LTB was used to compare with the *in vitro* lipid profile at the transcription level.
The TAG accumulation observed in hypoxic *Mtb* cells provided evidence that the oxygen-depletion model we employed resembled dormant-like bacilli [@pone.0058378-Daniel1]. *Mtb* under both, aerobic and NRP2 conditions, showed similar mycolate content, which is contrary to earlier studies that suggested that hypoxia is an important factor that induces the expression of proteins involved in mycolic acid biosynthesis [@pone.0058378-Starck1]. Thus, the changes in the cell wall lipids that we observed in NRP2 cultures might be at the level of non-covalently attached glycolipids. The observation that hypoxic *Mtb* H37Rv results in negative neutral red staining suggests impaired production of some methyl-branched glycolipids and a possible reorganisation of the mycobacterial cell wall [@pone.0058378-GonzaloAsensio1], [@pone.0058378-Cardona1]. SL, a methyl-branched glycolipid, is clearly decreased ([Figure 2](#pone-0058378-g002){ref-type="fig"}) in NRP2 stage**,** that contribute to the negative neutral red staining [@pone.0058378-Cardona1]. This behaviour is also observed in *phoP* mutants that are deficient in SL production [@pone.0058378-GonzaloAsensio1].
The transcriptional profile observed in hypoxia of genes involved in the SL and DAT/PAT biosynthesis may be produced as a response to the metabolic stress generated by gradual oxygen depletion. As it can be deduced from *Mtb mmpL8* knock-out [@pone.0058378-Brennan1], [@pone.0058378-Jackson1], the significantly low transcription of *mmpL8* avoids SL translocation to the mycobacterial cell wall and the subsequent complete biosynthesis of this glycolipid, which may partially explain the very low SL content in crude lipid extracts of NRP2-stage *Mtb* ([Figure 2](#pone-0058378-g002){ref-type="fig"} and [6](#pone-0058378-g006){ref-type="fig"}). In addition, the high *pks2*, *papA1* and *papA2* transcription in NRP2 suggests over-production of the hydroxyphtioceranic and phtioceranic acids that bind trehalose and induce the intracellular synthesis of diacylated sulfoglycolipids (Ac~2~SGL), precursor glycolipids that are more polar than the completely assembled SL [@pone.0058378-Layre1], [@pone.0058378-Converse1]; therefore, it is probably accumulated as a polar glycolipid.
{#pone-0058378-g006}
The increase of *pks3/pks4* transcription during hypoxia might induce the intracellular accumulation of mycolipenic and mycolipanoic acids. As observed in SL biosynthesis, the reduction of *mmpL10* transcription probably avoids the complete PAT biosynthesis and DAT translocation to the cell wall ([Figure 6](#pone-0058378-g006){ref-type="fig"}), which partially explains the DAT and polar glycolipids accumulation observed in lipid extracts from *Mtb* under NRP2-stage ([Figure 2B](#pone-0058378-g002){ref-type="fig"}); however, we cannot ruled out the possibility that other glycolipids can co-migrate with DAT in our chromatography study. Dormant-like mycobacteria use alternative carbon sources as fatty acids and cholesterol [@pone.0058378-MuozElias1], [@pone.0058378-Ouellet1]. Under hostile conditions, *Mtb* up-regulates the WhiB3 protein, which maintains redox homeostasis by regulating the biosynthesis of SL, TAG, PAT and phtiocerol dimycocerosate (PDIM) to reduce the metabolic stress generated by fatty acid catabolism under oxygen-limited conditions [@pone.0058378-Singh1]. Consequently, hydroxyphtioceranic, phtioceranic, mycolipenic and mycolipanoic acids, together with SL and PAT precursors, can be accumulated in NRP2; in this sense, an increased expression of *papA2* and *papA1* may be correlated with the accumulation of lipid intermediates used as energy source substrates [@pone.0058378-Daniel1], [@pone.0058378-MuozElias1].
Because the Wayne model does not establish the direct interaction of persistent mycobacteria with host cells, the present study was complemented with analysis of the transcriptional profile of SL and DAT/PAT using murine models. In the model of progressive TB, the low transcription of mycobacterial genes on day 1 is predictable because *Mtb* is in the process of adaptation. It appears that lipids such as SL and DAT/PAT are more relevant during late disease, i.e., after 28 days of infection, when in this model, there is a decrease in the protective Th-1 immune response, high bacilli loads and high transcription levels of genes involved in SL and PAT biosynthesis are observed ([Figure 5](#pone-0058378-g005){ref-type="fig"} and [S3](#pone.0058378.s003){ref-type="supplementary-material"}). Thus, it seems that the biological role of these glycolipids is more relevant during advanced disease when there is high bacilli proliferation and tissue damage (day 120) [@pone.0058378-Timm1], [@pone.0058378-HernndezPando2].
In the case of chronic infection similar to LTB, the gene transcriptional profile also shows that SL and PAT synthesis is incomplete, similar to that of the Wayne model of hypoxia. When confined in granulomas for a long time, *Mtb* has to face a predominant Th-1 immune response and high expression of TNF-α; therefore, biosynthesis of SL and PAT precursors may be associated with mechanisms of immune evasion. Furthermore, the inhibition of leukocyte migration by mycolipenic and mycolipanoic acid *in vitro* has been previously observed [@pone.0058378-Jackson1]. Thus, reduction in trehalose-based glycolipids production in LTB could be attributable to adaptation of tubercle bacilli to the hostile environment of granulomas; precursors of these glycolipids are may be preferentially used as a reserve energy and carbon source, inducing an increase in the transcription of *pks* and *papA* genes, as we observed [@pone.0058378-Daniel1], [@pone.0058378-MuozElias1], [@pone.0058378-Timm1]. *mmpL8* and *mmpL10* exhibit a tendency to increase as the course of long lasting infection progresses. This behaviour suggests that mycobacteria have a differential production of some trehalose-based glycolipids in the initial latent infection [@pone.0058378-Wayne2], [@pone.0058378-MuozElias1] as part of an adaptive process as seen in the *in vitro* NRP2 stage. In the reactivation stage, the transcriptional profile of genes is similar to that on day 120 of progressive infection in mice, suggesting that SL and PAT are involved in anti-inflammatory process related to active TB, such as pneumonic lesions and other tissue damage that are also promoted by the activity of the *rpfB* reactivation promoter [@pone.0058378-Hett1].
As expected, *phoP* shows a high transcription rate in hypoxia (NRP1 and NRP2) and during the entire process of latent infection, this observation was in agreement with the increase of the *pks* genes transcription [@pone.0058378-GonzaloAsensio2] and the subsequent accumulation of polar precursors. This behaviour is important because the virulence regulator *phoP* induces transcription of the global latency regulator Rv3133/*dos*R, together with the genes responsible for the initial and enduring hypoxic response and genes involved in the biosynthesis of the trehalose-based glycolipids that are correlated with *Mtb* virulence [@pone.0058378-Sherman1], [@pone.0058378-Rustad1].
In conclusion, the results obtained in this work suggest that LTB produces changes in trehalose-based glycolipid and long-chain fatty acid production that probably affect *Mtb* virulence. It is possible that latent *Mtb* uses SL and PAT precursors only as a metabolic adaptation to the granuloma environment, while in late active TB, these glycolipids are over-produced, likely because they play a role in virulence.
Supporting Information {#s5}
======================
######
***In vitro*** **transcription of genes in oxygenated and hypoxic** ***M. tuberculosis*** **H37Rv.** Transcription was measured during the exponential and stationary phases of growth and the non-replicative persistence 1 and 2 (NRP1 and NRP2) stages of hypoxia. The transcription of genes was normalised to the mean value of 16SrRNA. Asterisks indicate statistically significant differences compared to exponential phase values (*P*\<0.05).
(TIF)
######
Click here for additional data file.
######
***In vivo*** **transcription of genes in** ***M. tuberculosis*** **H37Rv during long-lasting TB infection in mice.** The transcription of genes was normalised to the mean value of 16SrRNA. Asterisks indicate statistically significant differences compared to transcription values at day 28 of progressive infection TB (*P*\<0.05).
(TIF)
######
Click here for additional data file.
######
***In vivo*** **transcription of genes in** ***M. tuberculosis*** **H37Rv during progressive tuberculosis in mice.** The transcription of genes was normalised to the mean value of 16SrRNA. Asterisks indicate statistically significant differences compared to transcription values at day 28 of progressive infection TB (*P*\<0.05).
(TIF)
######
Click here for additional data file.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
[^2]: Conceived and designed the experiments: JR AR LS JG-y-M RH CH CS. Performed the experiments: JR AR. Analyzed the data: JR JG-y-M RH CS. Contributed reagents/materials/analysis tools: JG-y-M RH CS. Wrote the paper: JR JG-y-M RH CS.
[^3]: Current address: Institute for Molecular Biosciences, Goethe University, Frankfurt, Germany
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Therapies for metastatic advanced thyroid cancers are inadequate with most therapeutic approaches failing to produce meaningful long-term responses (Smallridge et al. [@CR27]; Licitra et al. [@CR16]). The relatively modest response rates with tyrosine kinase inhibitors (TKIs) have led to several ongoing trials using combinations of TKIs with other targeted agents in an effort to improve the clinical outcome (Cohen et al. [@CR6]; Gupta-Abramson et al. [@CR12]; Sherman et al. [@CR25]; Schlumberger et al. [@CR22]; Bible et al. [@CR1]).
Histone deacetylase inhibitors (HDACIs) are a promising class of drugs in thyroid cancer; in vitro and in vivo studies of vorinostat (SAHA) (Luong et al. [@CR17]) showed that this agent is active against human thyroid cancer cell lines. However, data suggest that single-agent HDACIs may not be a viable treatment strategy in solid cancers (Ellis and Pili [@CR10]; Woyach et al. [@CR30]; Venugopal and Evans [@CR29]). As such, a number of trials are pursuing combinations of HDACIs with other agents both in thyroid cancer and other types of solid malignancies (Munster et al. [@CR19]). In this study, we tested two relatively new HDACIs (belinostat and panobinostat) and multiple TKIs alone and in combination against a panel of nine human thyroid cancer cell lines that have been validated as to their accuracy (Schweppe et al. [@CR23]); these HDACIs appear active both in vitro and in vivo (belinostat) and appear to have enhanced activity when combined with one of several TKIs.
Materials and methods {#Sec2}
=====================
Reagents {#Sec3}
--------
Belinostat and panobinostat (HDACIs) were obtained from National Cancer Institute (NCI, USA) and Selleck Chemicals (Boston, MA), respectively. Pazopanib, motesanib, sorafenib and dasatinib (TKIs) (LC Laboratories) were dissolved in DMSO and stored in aliquots at −20 °C until use. Antibodies to acetylated histone H3 (AcH3), p21^Waf1^, PARP, pAKT (Ser473), pERK, AKT, ERK and GAPDH were obtained from Cell Signaling Technology (MA, USA).
Cell lines and culture {#Sec4}
----------------------
Cal62 (anaplastic---KRAS G12R mutant), SW1736 (anaplastic---BRAF mutant), T238 (anaplastic---PI3K and BRAF mutant), Hth7 (anaplastic---NRAS Q61R), T241 (unknown---no mutation), T351 (unknown---no mutation), C643 (anaplastic---HRAS G13R), BHP2-7 (papillary---RET/PTC rearrangements) and Hth83 (unknown---HRAS G13R) were generously provided by Dr. James A. Fagin (Memorial Sloan Kettering Cancer Center, New York, USA). Hth7 was grown in DMEM with 10 % fetal bovine serum (Hyclone, Logan, UT). The rests of the cell lines were maintained in RPMI (Life Technologies, Carlsbad, CA) containing 10 % fetal bovine serum, 1 % penicillin--streptomycin and 1 % glutamine with the exception of SW1736, which was supplemented with 1 % MEM non-essential amino acid. All cells were grown at 37 °C and 5 % CO~2~.
Cytotoxicity assay {#Sec5}
------------------
For measurement of cell viability, 1.0--5.0 × 10^3^ cells were plated in 96-well flat-bottomed microtiter plates and grown overnight at 37 °C. The following day, the cells were treated with 10 μl of various concentrations of either HDAC inhibitors/TKI/combination or vehicle and incubated for 48/72 h. After 48/72 h, 250 μg of MTT \[3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide\] (Sigma-Aldrich) was added to each well and incubated at 37 °C for 2--4 h, after which 100 μl of DMSO was added, as previously reported (Chan et al. [@CR4]). Absorbance was measured with a Tecan Infinite 200 PRO spectrophotometer (Mannedorf, Switzerland) at a wavelength of 595 nm. As previously mentioned, the percentage viability at each drug concentration was calculated using the following formula:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ \frac{{({\text{Absorbance}}\;{\text{of}}\;{\text{treated}}\;{\text{cells}} - {\text{Absorbance}}\;{\text{of}}\;{\text{media}})}}{{({\text{Absorbance}}\;{\text{of}}\;{\text{untreated}}\;{\text{cells}} - {\text{Absorbance}}\;{\text{of}}\;{\text{media}})}} $$\end{document}$$The percentage values obtained were then plotted using GraphPad Prism^®^ software nonlinear regression (curve fit) to obtain the half maximal inhibitory concentration (IC~50~) and its 95 % confidence interval (CI).
The dose ranges tested: belinostat (100--10,000 nM); panobinostat (1--2,000 nM); dasatinib (1--300 nM); sorafenib, pazopanib and motesanib (100--5,000 nM).
Assessment of synergism by the median effects model of Chou-Talalay (Calcusyn^®^) {#Sec6}
---------------------------------------------------------------------------------
The combinations of HDACIs and TKIs were assessed by the median effects model of Chou-Talalay using Calcusyn^®^ software from BioSoft^®^. The interpretation of the normalized isobolograms and interpretation of the degree of synergism is as per the combination index (CI) stated in Supporting Information Fig. S1.
Gene expression profiling and analysis {#Sec7}
--------------------------------------
Using Affymetrix Human Gene 1.0ST array (Affymetrix, Santa Clara, CA), mRNA expression profiling was done in either biological triplicates (for BHP control and BHP panobinostat 100 nM, 30 h) or biological duplicates (BHP belinostat 50 μM, 30 h). Total RNA was reversed transcribed, amplified and labeled according to manufacturer's protocol. Labeled cDNA was hybridized onto Human Gene 1.0ST array for 16 h at 60 rpm in a GeneChip Hybridization Oven 640 (Affymetrix). Following manufacturer's instructions, the arrays were washed and scanned using GeneChip Scanner 3000 7G. The CEL files were imported into GeneSpring Software V11 for data normalization and identification of differentially expressed mRNAs of interest. As mentioned previously, the CEL files were preprocessed using RMA16 algorithm and CORE transcript level. Data were quantile normalized, and those expression values less than the 20th percentile were filtered out. A list of differentially expressed mRNAs of interest was generated using the fold change function (Chan et al. [@CR4]). Using hierarchical clustering, heatmaps were generated and gene ontology analysis was done using GO. To understand further the biological functions of the differentially expressed genes, Web-based software, MetaCore (GeneGo, St. Joseph, MI), was used to perform gene ontology and pathway/network analysis as mentioned previously (Chan et al. [@CR4]). Several algorithms to enable both the construction and analysis of gene networks were integrated as previously described (Ekins et al. [@CR9]).
Western blotting {#Sec8}
----------------
Lysates were collected from cells treated with HDAC inhibitor for 30 h. Briefly, cells were lysed at 4 °C in RIPA lysis buffer, and the insoluble material was removed by centrifugation at 4 °C, 10,000*g* for 10 min. Proteins in the cell lysates were quantified by Bradford Assay. Forty micrograms of protein from each sample was resolved by SDS-PAGE and electroblotted to Immobilon-P Transfer Membranes (Millipore, Billerica, MA). Membranes were probed with the various primary antibodies from Cell Signaling Technologies. Following binding of the primary antibodies, horseradish peroxidase (HRP)-conjugated anti-rabbit/anti-mouse secondary antibodies were introduced to the membrane. Bound secondary antibody was detected by using Amersham ECL Plus Western blotting detection reagents (GE Healthcare, USA). Membranes were stripped and probed for GAPDH as loading controls or the non-phosphorylated protein (where applicable) (Chan et al. [@CR4]).
Apoptosis assay {#Sec9}
---------------
According to the instructions from the manufacturer, 10^5^ cells were treated with either belinostat or panobinostat for 48 h and labeled with FITC-conjugated annexin V antibody and propidium iodide (PI) using annexin V--FITC Apoptosis Detection kit I (BD Biosciences, San Jose). Positive cells were detected by fluorescence-activated cell sorting as mentioned previously (Chan et al. [@CR4]).
Murine studies {#Sec10}
--------------
All mice were fed a standard chow diet. Six-week-old female athymic nu/nu mice---NCRFU---were purchased from Taconic Hudson (NY). Upon arrival, these animals were quarantined for 1 week in the animal facility and maintained in a 12 h light/dark cycle, with free access to food and water.
Mice were anaesthetized with inhalational isoflurane (Hospira Lakeforest, IL). BHP2-7 cells (2 × 10^7^) were resuspended in 50 μl Matrigel (BD Biosciences) and 50 μl PBS for each tumor. Cells were injected subcutaneously on both flanks of the immunodeficient mice. A total of 32 tumors were inoculated (8 in control and 8 in drug treatment mice). Mice were injected with belinostat intraperitoneal (Cheung et al. [@CR5]) (100 mg/kg/injection), 5 days a week for 3 weeks starting on day 3. The weight of the mice was measured every week. At the end of either drug or vehicle-only treatment (52 days after tumor cell injection), animals were killed by CO~2~ inhalation, tumor dissected and volumes calculated according to the formula: (length × width × thickness × 0.5236) (Chan et al. [@CR4]). During the experiment, external measurements were made in millimeters using vernier callipers. At completion of the study, tumors were formalin-fixed and processed for histological and immunohistochemical (IHC) analyses. Murine studies were approved by the Animal Care and Use Committee of Cedars-Sinai Medical Center, and all animal care was in accordance with the IACUC guidelines.
Results {#Sec11}
=======
Cytotoxicity assay {#Sec12}
------------------
To assess the anti-proliferative activity of either the HDACIs or TKIs, dose--response studies were performed using MTT assay. Mean proportion viability of the cells at each drug concentration was plotted using GraphPad Prism software^®^, and the drug dose that inhibited 50 % growth (relative to vehicle-treated control) \[IC~50~\] at 72 h of culture was obtained from the software (nonlinear regression curve fit). For range of drug testing, the pharmacologically relevant doses of each of the drugs were inferred from phase I clinical trials (see "[Materials and methods](#Sec2){ref-type="sec"}") (Giles et al. [@CR11]; Hurwitz et al. [@CR14]; Minami et al. [@CR18]; Demetri et al. [@CR7]; Schlumberger et al. [@CR22]; Steele et al. [@CR28]). Both belinostat and panobinostat reduced thyroid cancer cell viability in vitro in 8 of the 9 cell lines tested at a pharmacologically achievable dose; the PI3K + BRAF mutant (T238) cell line was resistant to panobinostat (Table [1](#Tab1){ref-type="table"}). Only sorafenib \[8 of 9 cell lines (IC~50~ 5.0--13 μM)\], pazopanib \[1 of 9 cell lines (IC~50~ 3.4 ± 0.2 μM)\] and dasatinib \[2 of 9 cell lines, (IC~50~ 47 ± 3 and 35 ± 5 nM)\] were effective among the TKIs (Table [1](#Tab1){ref-type="table"}).Table 1Dose of drug that inhibited 50 % growth of human thyroid cancer cell line in vitroCell lineMutation/arrangementsIC50 (nM) BelinostatIC50 (nM) PanobinostatIC50 (nM) DasatinibIC50 (μM) SorafenibIC50 (nM) PazopanibIC50 MotesanibCal62KRAS470 ± 3633 ± 435 ± 513 ± 1.1NRNRHth7NRAS400 ± 5615 ± 2NR6.8 ± 0.8NRNRHth83HRAS720 ± 10034 ± 5NR11 ± 1.0NRNRC643HRAS970 ± 12071 ± 10NR6.5 ± 0.9NRNRSW1736BRAF380 ± 4035 ± 8NRNRNRNRT241Nil750 ± 6565 ± 7NR7.5 ± 0.5NRNRT351Nil1,100 ± 17050 ± 10NR11 ± 2.0NRNRBHP2-7RET/PTC770 ± 6737 ± 647 ± 35.0 ± 0.8340 ± 20NRT238PI3 K/BRAF6,800 ± 7401,500 ± 200NR6.5 ± 1.2NRNRTabulated results of HDACI (belinostat and panobinostat) and tyrosine kinase inhibitor (TKI dasatinib, sorafenib, pazopanib and motesanib)-treated thyroid cancer cell lines at 72 h of drug exposure. Mean 50 % inhibitory concentration (IC50 ± SD) of experiments done in triplicate experiments with each experiment done with triplicate wells using MTT assay. Nil: No mutations detected. NR: IC50 not reached (i.e., pharmacologically relevant IC50 not reached). See "[Materials and methods](#Sec2){ref-type="sec"}" for details
Belinostat- and panobinostat-treated BHP2-7 thyroid cancer cell line has overlapping mRNA changes {#Sec13}
-------------------------------------------------------------------------------------------------
In order to gain insights into which common pathways is altered by the two HDACIs, BHP2-7 cells were exposed to either belinostat 50 μM or panobinostat 100 nM for 30 h. RNA was extracted and subjected to microarray analysis (Supporting Information Table S1). Only genes with ≥twofold change (panobinostat treatment) or ≥fourfold change (belinostat treatment) were deemed to be significantly altered. Among the 528 and 573 altered genes for panobinostat and for belinostat, respectively, 153 of these genes were commonly altered. These 153 genes constituted 3 main pathways: cell cycle-related, DNA damage and apoptosis (Supporting Information Fig. S2) which prompted additional studies.
Belinostat and panobinostat treatment-related protein changes {#Sec14}
-------------------------------------------------------------
Both belinostat and panobinostat caused acetylation of histone H3 and induced the expression of p21^Waf1^ in four representative thyroid cancer cell lines (Cal62, SW1736, T238 and BHP2-7) irrespective of their mutational composition (Fig. [1](#Fig1){ref-type="fig"}, Panel a). Of note, T238 cells were relatively resistant to growth inhibition by the HDACIs and had minimal induction of p21. Protein changes in the RAS-RAF-MEK-ERK and PI3K-AKT-mTOR pathways due to HDACIs were examined as these are important signaling proteins in thyroid cancer. Reduced levels of both pERK and pAKT were observed in all four thyroid cancer lines with both belinostat and panobinostat treatments (Fig. [1](#Fig1){ref-type="fig"}, Panel b). Also, both HDACIs caused apoptosis as measured by cleaved PARP (Fig. [1](#Fig1){ref-type="fig"}, Panel c). \[No cleaved PARP was detectable in resistant T238 cultured with panobinostat 100 nM.\]Fig. 1Immunoblots of thyroid cancer cell lines (BHP2-7, SW1736, T238 and Cal 62) treated with either belinostat 50 μM (Bel), panobinostat 100nM (Pan) or diluent control (Cont) (30 h) and analyzed for expression levels of various proteins. (**a**) Phosphorylated pAKT and pERK. (**b**) And cleaved PARP (*arrowhead*)
Annexin V/propidium iodide assay {#Sec15}
--------------------------------
Consistent with the detection of cleaved PARP observed on Western blot, belinostat (50 μM) and panobinostat (100 nM) (30 h) induced early and late apoptosis in the BHP2-7, Cal62 and SW1736 thyroid cancer cell lines as measured by annexin V/propidium measurements (Table [2](#Tab2){ref-type="table"}). Panobinostat (100 nM) resulted in less apoptosis (11--37 %) than belinostat (50 μM) (43--68 %). Consistent with the Western blot data, the T238 cell line had fewer apoptotic cells after culture with either panobinostat (3 %) or belinostat (24 %) as compared to the other HDACI-treated cell lines (Table [2](#Tab2){ref-type="table"}).Table 2Annexin V/propidium iodine assayControl (apoptosis) (%)Panobinostat (100 nM) (apoptosis) (%)Belinostat (50 μM) (apoptosis) (%)Cal6233768SW173631156BHP2-751643T2382324Combined early and late apoptosis results \[annexin V early and propidium iodine (late) positive measured by FACS\] (expressed as percentage of total number of cells) tabulated for BHP2-7, Cal62, SW1736 and T238 cell lines after 30-h culture. Results represent the mean of triplicate cultures. Means did not vary more than 10 % between experiments
Human thyroid cancer xenograft study {#Sec16}
------------------------------------
Belinostat was injected intraperitoneally into immunodeficient mice carrying BHP2-7 xenografts (100 mg/kg/day, 5 days/week, 52 days). Prominent inhibition of growth of the tumors occurred in the experimental mice compared to control mice receiving diluent alone (Fig. [2](#Fig2){ref-type="fig"}a). Experimental mice did not lose body weight (Fig. [2](#Fig2){ref-type="fig"}b) and appeared and behaved normally during the study.Fig. 2Effect of belinostat on proliferation of human thyroid cancer xenografts growing in athymic mice. Graph of mean xenograft tumor volume (mm^3^) in belinostat-treated (100 mg/kg/5 day per week, I.P.) versus vehicle control-treated (treated similarly) mice over 52 days. Sixteen tumors (BHP2-7) were in both groups. (**a**) Graph of mean weights (in grams) of mice in the belinostat-treated and control-treated groups over the duration of experiment (**b**)
TKIs and their combination with either belinostat or panobinostat with TKIs {#Sec17}
---------------------------------------------------------------------------
Combinations of either belinostat or panobinostat with one of the TKIs (see Table [1](#Tab1){ref-type="table"}) were tested against 3 thyroid cancer cell lines and assessed by MTT assay and Calcusyn^®^ analysis (Table [3](#Tab3){ref-type="table"}a--c). HDACI--sorafenib combinations were antagonistic, while motesanib-HDACI combinations were only additive (data not shown). However, either belinostat or panobinostat with either dasatinib or pazopanib was synergistic in their anti-proliferative activity against BHP2-7 (RET/PTC rearrangement), Cal62 (KRAS mutant) and SW1736 (BRAF mutant) thyroid cancer cells. The degree of synergism varied among the above combinations but was strongest for the Cal62 cell line \[i.e., combination index (CI) was lowest\] (interpretation of the combination index appears on Supporting Information Fig. S1).Table 3Combination index of (a) belinostat/panobinostat and dasatinib on SW1736 cells, (b) belinostat/panobinostat and dasatinib on Cal 62 cells, (c) belinostat/panobinostat and pazopanib on BHP 2-7 cellsDasatinib (nM)Belinostat (nM)Panobinostat (nM)(a)2505007501,00012040501000.470.410.390.290.140.480.540.56Dasatinib (nM)Belinostat (nM)Panobinostat (nM)(b)2505007501,00081624451000.130.140.150.140.170.160.180.29Pazopanib (μM)BelinostatPanobinostat1 μM30 nM10.660.5020.630.5130.520.4740.54--Combination index described more fully in Fig. S1. Values ≤0.90 are consistent with synergism, and the lower the value, the greater the synergism
Discussion {#Sec18}
==========
We found that belinostat and panobinostat inhibited growth in vitro, induced apoptosis and dephosphorylated AKT and ERK in a panel of thyroid cancer cell lines; and belinostat remarkably inhibited tumor growth in mice carrying thyroid cancer xenografts. Studies have shown that this family of drugs can have a myriad of effects (Dokmanovic et al. [@CR8]). As their name implies, they can reverse the activity of HDAC, causing acetylation of histones which opens the chromatin to allow transcription of a large array of genes. Furthermore, these compounds can acetylate numerous proteins resulting in change of their function (Bolden et al. [@CR2]). Our microarray analysis of BHP2-7 treated with either belinostat or panobinostat showed modulation of genes involved in pathways associated with the cell cycle, DNA damage and apoptosis-related processes.
The HDACIs acetylated H3 protein in the four representative cell lines, including T238 which were resistant to growth inhibition by HDACI (especially panobinostat) (Fig. [1](#Fig1){ref-type="fig"}). These cells had minimal induction of p21, suggesting that this cyclin kinase inhibitor may potentially represent a biomarker of growth inhibition by the HDACI. Various in vitro studies of MAPK and PI3K/AKT inhibitor monotherapy have shown conflicting results in thyroid cancer (depending on mutational status and growth conditions) (Schweppe et al. [@CR24]). However, combinations of agents blocking both the PI3K-AKT-mTOR and MAPK pathways show greater anti-proliferative activity (Jin et al. [@CR15]) and are now in clinical trials (e.g., combinations of sorafenib and everolimus) \[[www.clinicaltrial.gov](http://www.clinicaltrial.gov)\]. We show that both belinostat and panobinostat markedly inhibited activation of both of these pathways in the four representative thyroid cancer cell lines. But T238 cells were resistant to growth inhibition by the HDACIs suggesting that inhibition of either phosphor-AKT or phosphor-ERK does not assure that the HDACI will be growth inhibitory.
One notable characteristic of thyroid cells is the ability to exchange extracellular iodine (I) for intracellular sodium (Na) by way of the Na/I symporter. The intracellular iodine has to be organified for it to stay in the cell rather than efflux out. While well-differentiated metastatic thyroid cancer can be treated with ^131^I which is selectively taken up by the thyroid Na/I symporter, radioiodine-resistant thyroid cancers are not amenable to this form of therapy. We found that belinostat and panobinostat could induce both Na/I symporter mRNA and protein in thyroid cancer cells, but the levels of intracellular ^131^I accumulation were low even though TPO was induced by these HDACIs (data not shown).
In recent years, phase II clinical trials of various small molecule tyrosine kinase inhibitors (including sorafenib, motesanib and pazopanib) (Gupta-Abramson et al. [@CR12]; Sherman et al. [@CR25]; Bible et al. [@CR1]) have shown modest activity against thyroid cancers (mainly stabilization disease by RECIST criteria) (Cabanillas et al. [@CR3]). The exact mechanism of action of these agents is still uncertain given the poor correlation between the predicted in vitro targets (e.g., RET and BRAF inhibitors) of these agents, mutational status of the tumors (RET and BRAF mutations) and clinical response. Of note, the RECIST clinical response rates of the above mentioned kinases are all remarkably similar in spite of the differences in spectrum of their kinase targets. The lack of efficacy of the tested TKIs is not surprising given our previous study (Chan et al. [@CR4]) which showed that most thyroid cancer cell lines are either resistant to a large number of TKIs or else are effective only at high IC~50~ which are not pharmacologically relevant. A prior in vitro and in vivo study of sorafenib (Salvatore et al. [@CR21]) showing that it was effective against a panel of BRAF V600E thyroid cancer cell lines is now known to have been confounded by the fact that most of the tested cell lines were not thyroid in origin (Schweppe et al. [@CR23]) and that sorafenib is not a potent BRAF inhibitor.
The sensitivity of BHP2-7 (RET/PTC rearrangement) to pazopanib as well as the sensitivity of BHP2-7 and the KRAS mutant Cal62 cell line to dasatinib within the pharmacological range merits further study. The lack of efficacy of sorafenib against SW1736 (BRAF mutant) that we observed is now known to be due to the fact that sorafenib does not have potent anti-BRAF properties (Smalley and Flaherty [@CR26]). Trials of PLX4032 in thyroid cancer, a proven potent BRAF inhibitor in melanoma, have recently been shown to have poor activity against RAS or RET mutants (Salerno et al. [@CR20]). This is partly due to the paradoxical ability of Raf inhibitors to stimulate wild-type Raf to activate the MAPK pathway and enhance growth of these tumors (Hatzivassiliou et al. [@CR13]).
The synergistic anti-proliferative activity of either belinostat or panobinostat combined with either pazopanib or dasatinib in the thyroid cancers carrying either RET/PTC1 rearrangement (BHP2-7), KRAS G12R mutation (Cal62) or BRAF V600E mutation (SW1736) is of interest. The mutations in these three lines are important components in the MAPK cascade. We found that the downstream pathways of these oncogenes could be inhibited by the HDACI, and thus, the TKI and HDACI may work in unison resulting in their anti-proliferative activity.
In summary, belinostat and panobinostat were effective against a wide range of thyroid cancer cell lines irrespective of their mutational composition (with the possible exception of the PI3K mutant, T238 cell line). This, in concert with the activity of belinostat to suppress the growth of thyroid cancer xenografts in athymic nude mice with no obvious side effects, suggests that clinical studies with these HDACs should be pursued. Furthermore, the synergistic combination of HDAC inhibitors-TKIs has previously not been reported and might present a clinical opportunity.
Electronic supplementary material
=================================
{#Sec19}
######
Supplementary material 1 (TIFF 116 kb)
######
Supplementary material 2 (TIFF 473 kb)
######
Supplementary material 3 (DOCX 22 kb)
Glenn D. Braunstein and H. Phillip Koeffler share last authorship.
This work was funded by the Singapore Ministry of Health's National Medical Research Council under its Singapore Translational Research (STaR) Investigator Award to H. Phillip Koeffler and NIH Grants R01CA026038-32, U54CA137785-01. This study is dedicated to the memory of Professor David Golde, a mentor and friend.
Conflict of interest {#d30e1391}
====================
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Adenosine 3′,5′-cyclic monophosphate (cAMP) is an ubiquitous second messenger which directly activates protein kinase A (PKA) and EPACs (exchange proteins directly activated by cAMP) and opens cyclic nucleotide-gated channels \[[@B1], [@B2]\]. PKA is the primary downstream effector of cAMP, regulating neurotransmitter release through activation of Ca^2+^ channels or inactivation of K^+^ channels \[[@B3], [@B4]\]. cAMP is also implicated in memory and synaptic plasticity at the hippocampus through activation of EPACs and through PKA-mediated CREB (cAMP responsive element binding protein) activation \[[@B5], [@B6]\]. The cAMP signal is transitory and regulated through the opposing actions of adenylyl cyclase and phosphodiesterases \[[@B1]\].
The G~i/o~-protein coupled cannabinoid CB~1~ receptors and adenosine A~1~ receptors are both expressed at high levels in the hippocampus \[[@B7], [@B8]\], where they inhibit adenylyl cyclase and consequently decrease cAMP production \[[@B9], [@B10]\]. Furthermore, both receptors colocalize in hippocampal CA3 pyramidal neuron axon terminals, in which they inhibit glutamatergic synaptic transmission to CA1 pyramidal neurons \[[@B11]--[@B13]\], are involved in impairment of learning and memory \[[@B14], [@B15]\], protect against neurotoxic insults, and have antinociceptive action \[[@B16]--[@B19]\].
Given the similarity between the transducing pathways operated by adenosine A~1~ and cannabinoid CB~1~ receptors, clarification of the combined activity of these receptors is a particularly interesting issue since both receptors are targets for widely consumed drugs, such as caffeine, an adenosine receptor antagonist, and the psychotropic Δ^9^-tetrahydrocannabinol (THC), a cannabinoid CB~1~ receptor agonist \[[@B20]\]. Interaction between A~1~ and CB~1~ receptors has been reported in*in vivo* studies, where an adenosine A~1~ receptor-mediated enhancement of cannabinoid CB~1~ receptor-induced impairment of short-term spatial memory and motor incoordination were observed \[[@B20], [@B21]\]. However, the interactions observed*in vivo* might be polysynaptic and dependent on circuitry, not necessarily reflecting receptor interaction at the cellular and molecular levels. Previous studies indicate that when acutely coactivated, adenosine A~1~ and cannabinoid CB~1~ receptors independently inhibit excitatory synaptic transmission in the rat hippocampus and additively stimulate G-protein activation in brain membranes (\[[@B12], [@B22]\], but see \[[@B23]\]).
Since the putative independence of the acute inhibitory effect of adenosine A~1~ and cannabinoid CB~1~ receptors could be a localized phenomenon, restricted to excitatory synaptic transmission in CA1 area of hippocampus \[[@B12]\], we now further investigated if it also applies to second messenger formation in the whole hippocampus. For that purpose, we studied how the acute coactivation of A~1~ and CB~1~ receptors modulates adenylyl cyclase activity in rat hippocampal slices. Clarification of the combined activity of these receptors on cAMP production would also help to understand how cells integrate signals triggered from both A~1~ and CB~1~ receptors to regulate brain cells activity. On the other hand, since even subchronic activation of A~1~ receptor can induce its desensitization \[[@B24]\] and therefore might cause cross desensitization of the CB~1~ receptor \[[@B25]\], the effect of short-term adenosine A~1~ receptor desensitization on the combined action of adenosine A~1~ and cannabinoid CB~1~ receptors was also investigated.
2. Material and Methods {#sec2}
=======================
2.1. cAMP Accumulation in Hippocampal Slices {#sec2.1}
--------------------------------------------
The experiments were performed using acute hippocampal slices taken from young adult male Wistar rats (6--8 weeks old). The animals were handled according to European Community guidelines and Portuguese law concerning animal care and were anesthetized with halothane before decapitation. The brain was rapidly removed and transferred to ice-cold Krebs-Henseleit buffer with the following composition (mM): NaCl 118, KCl 4.7, KH~2~PO~4~ 1.2, MgSO~4~ 1.2, CaCl~2~ 1.3, NaHCO~3~ 25, glucose 11.6, gassed with 95% O~2~ and 5% CO~2~ (pH 7.4). The brain was cut longitudinally, the two hippocampi were dissected and cross chopped (350 × 350 *μ*m) with a McIlwain tissue chopper. Sliced hippocampi were then placed in an Erlenmeyer, dispersed, and washed twice with buffer. The cross chopped hippocampal slices were transferred into a conical-bottom polypropylene tube and 50 *μ*L aliquots of gravity-packed slices (1-2 mg protein) were pipetted into flat-bottom propylene tubes (1.65 × 9.5 cm, 20 mL capacity) containing Krebs buffer and preincubated for 30 min at 37°C in a shaking (1 cycle*·*s^−1^) water bath. Since basal intracellular levels of cAMP in hippocampal slices are low and hard to quantify, most experiments were performed in the presence of forskolin and rolipram in order to increase cAMP concentration. Forskolin directly stimulates adenylyl cyclase while rolipram inhibits phosphodiesterase 4, the main enzyme responsible for cAMP degradation in the brain \[[@B26]\]. Incubation with drugs started with addition of rolipram (50 *μ*M final concentration). Forty-five min after rolipram addition, incubation proceeded in the absence or in the presence of forskolin (10 *μ*M) for a further 15--35 min period. Assays performed in the presence of forskolin, and its controls, also contained ethanol (0.02%, v/v), its vehicle. When used, WIN55212-2 (0.3--30 *μ*M), adenosine deaminase (2 U/mL), or DPCPX (50 nM) were present simultaneously with the start of incubation with rolipram, while AM251 (10 *μ*M) was added 30 min before addition of rolipram. CPA (10--150 nM final concentration), when present, was added 30 min after rolipram addition. In one set of experiments WIN55212-2 (30 *μ*M) was present since 5 h and 15 min before rolipram addition and in another set CPA (100 nM final concentration) was added 50 min before rolipram. The final volume after all drug additions was 300 *μ*L. Note that, usually, longer incubation times were used when testing the effect of WIN55212-2 than when testing the CPA effect; this was necessary because WIN55212-2 is very lipophilic and therefore needed longer incubations times to equilibrate with hippocampal slices and produce its inhibitory effect (see \[[@B12]\]). In fact we have found, in a previous electrophysiological study using hippocampal slices \[[@B12]\], that WIN55212-2 started to produce effect on neurotransmission only after 30 min after its application to the hippocampal slice, and it took 60--90 min to produce its maximal effect. When testing the effect of a drug, a parallel control assay was done in which a same volume of vehicle replaced the volume of drug solution added to the tube. Tubes were gassed for 20 s and capped, after slices or drug addition.
Incubations were stopped by adding 100 *μ*L of perchloric acid (HClO~4~, 10% w/v) solution containing EDTA (20 mM). Samples were sonicated for 2 minutes, placed on ice for 30 minutes, neutralized by addition (100 *μ*L) of potassium carbonate (K~2~CO~3~, 0.5 M), and vortexed for 2 minutes, allowing the CO~2~ to escape. The tubes were then placed on ice for an additional 15 minutes period to precipitate potassium perchlorate. The samples were centrifuged (5000 g, 10 min at 4°C) and 200 *μ*L aliquots, per sample, of the supernatants were collected and stored at −80°C for cAMP content analysis. The pellets were digested with NaOH (1 M) for 1.5 h at 37°C, neutralized, and individually assayed in duplicate for protein content by the method of Peterson \[[@B27]\]. The samples were analyzed for cAMP content using an enzyme immunoassay (EIA) kit (Cayman Chemical). cAMP concentration in each sample was expressed as pmol per mg of protein.
2.2. Drugs Used {#sec2.2}
---------------
(R)-(+)-\[2,3-Dihydro-5-methyl-3-(4-morpholinylmethyl)pyrrolo\[1,2,3-de\]-1,4-benzoxazin-6-yl\]-1-naphthalenylmethanone mesylate (WIN55212-2), 8-cyclopentyl-1,3-dipropylxanthine (DPCPX), N6-cyclopentyladenosine (CPA), N-(piperidin-1-yl)-5-(4-iodophenyl)-1-(2,4-dichlorophenyl)-4-methyl-1H-pyrazole-3-carboxamide (AM251), rolipram, and forskolin were purchased from Tocris (Bristol, UK). Adenosine deaminase was obtained from Sigma-Aldrich. Stock solutions of WIN55212-2 (20 mM), rolipram (20 mM), DPCPX (50 *μ*M), and AM251 (5 mM) were prepared in dimethyl sulfoxide (DMSO). Forskolin (50 mM) and CPA (2 mM) stock solutions were prepared in ethanol and water, respectively. Suitable dilutions of each stock solution with Krebs buffer were made before performing the experiments.
2.3. Data Analysis {#sec2.3}
------------------
The values are expressed as mean ± S.E.M. from *n* experiments. The significance of the differences between the mean values obtained in two different conditions, or when comparing means with zero, was evaluated by Student\'s *t*-test, where the paired Student\'s *t*-test was used whenever evaluating the significance of differences between two conditions tested in a paired way in the same experiment. When more than two different conditions were simultaneously compared, One-way ANOVA was used followed by the LSD post-hoc test. The maximal effect (*E* ~max~) and the concentration of agonist producing half-*E* ~max~ (EC~50~) were calculated by fitting the agonist concentration-response curve data to a Michaelis-Menten type equation, through nonlinear regression using the SPSS for Windows program version 16.0 (SPSS Inc., Chicago, Illinois, USA).
3. Results {#sec3}
==========
3.1. Maximal Effect, Potency, and Specificity of Adenosine A~1~ and Cannabinoid CB~1~ Agonists on Forskolin-Stimulated cAMP Accumulation {#sec3.1}
----------------------------------------------------------------------------------------------------------------------------------------
In the presence of rolipram (50 *μ*M), the cAMP accumulation was 40 ± 11 pmol/mg protein (*n* = 3), whereas the further addition of 10 *μ*M forskolin increased basal cAMP accumulation by about fivefold (to 202 ± 46 pmol/mg protein, *n* = 3, *P* \< 0.05, paired Student\'s *t*-test).
As shown in Figures [1(a)](#fig1){ref-type="fig"} and [1(b)](#fig1){ref-type="fig"}, both the adenosine A~1~ receptor selective agonist CPA (10--150 nM) and the cannabinoid CB~1~ receptor agonist WIN55212.2 (0.3--30 *μ*M) dose-dependently inhibited forskolin-stimulated cAMP accumulation in the hippocampus. Computerized curve fitting to the data shown in [Figure 1(a)](#fig1){ref-type="fig"} gave an EC~50~ for CPA of 35 ± 19 nM and a maximal decrease of cAMP accumulation (*E* ~max~) of 29% ± 5%, whereas for WIN55212-2 ([Figure 1(b)](#fig1){ref-type="fig"}) an EC~50~ of 6.6 ± 2.7 *μ*M and an *E* ~max~ of 31% ± 2% were obtained. Application of CPA (100 nM) caused a 21% ± 3% (*n* = 8) inhibition of cAMP accumulation, while when WIN55212-2 (30 *μ*M) was applied, the cAMP accumulation was decreased by 25% ± 4% (*n* = 9). We found these concentrations adequate to test the combined effect of CPA and WIN55212-2 on cAMP accumulation, since with them we obtained a robust effect. The inhibitory effect of CPA (100 nM) on cAMP accumulation was fully blocked by the adenosine A~1~ receptor selective antagonist DPCPX (50 nM; [Figure 2(a)](#fig2){ref-type="fig"}), while the inhibitory effect of WIN55212-2 (10 *μ*M) was strongly attenuated by the cannabinoid CB~1~ receptor selective antagonist AM251 (10 *μ*M; [Figure 2(b)](#fig2){ref-type="fig"}). Note that in the presence of AM251, WIN55212-2 produced a residual inhibitory effect on cAMP accumulation (4.3 ± 0.6, *n* = 3; [Figure 2(b)](#fig2){ref-type="fig"}). The choice of a 10 *μ*M concentration of WIN55212-2 when studying the reversal of its effect by AM251 was determined by the solubility of AM251. Since AM251 is very lipophilic, it is difficult for it to diffuse into the bulk of the slice so that it reaches the right concentration to efficiently inhibit cannabinoid CB~1~ receptors. Thus, the appropriate concentration of AM251, which depends on its low solubility in aqueous buffer, required that the concentration of WIN55212-2 would not surpass 10 *μ*M. Accumulation of cAMP was not affected by either DPCPX or AM251 alone ([Figure 2](#fig2){ref-type="fig"}).
3.2. Combined Effect of Adenosine A~1~ and Cannabinoid CB~1~ Agonists {#sec3.2}
---------------------------------------------------------------------
When CPA (100 nM) and WIN55212-2 (30 *μ*M) were applied together, respectively, 15 min and 45 min before forskolin, the combined application of WIN55212-2 and CPA produced a higher inhibition of cAMP accumulation (41% ± 6%) than that produced by either WIN55212-2 or CPA alone ([Figure 3(a)](#fig3){ref-type="fig"}). Furthermore, the combined effect of CPA and WIN55212-2 did not differ from the sum of the individual effects of each agonist (43% ± 8%; *P* \> 0.7, paired Student\'s *t*-test, [Figure 3(a)](#fig3){ref-type="fig"}).
3.3. CB~1~ Activity Remains Unaffected by Short-Term Desensitization of Adenosine A~1~ Receptors {#sec3.3}
------------------------------------------------------------------------------------------------
As we may observe in [Figure 3(b)](#fig3){ref-type="fig"}, increasing the preincubation period with CPA, from 15 to 95 min before forskolin addition, caused a significant (*P* \< 0.05) attenuation of the CPA effect on forskolin-stimulated cAMP accumulation in the hippocampal slice. In fact, when CPA (100 nM) was applied 95 min before forskolin, no significant effect of CPA was observed (*P* \> 0.22, Figures [3(b)](#fig3){ref-type="fig"} and [3(c)](#fig3){ref-type="fig"}) suggesting that short-term desensitization mechanisms were operating on A~1~ receptors. Consequently the possibility that adenosine A~1~ receptor desensitization could cross desensitize cannabinoid CB~1~ receptors and modify the cannabinoid CB~1~-mediated action on cAMP accumulation was investigated. After inducing short-term desensitization of A~1~ receptors by 95 min exposure to CPA, the inhibitory effect of WIN55212-2 (30 *μ*M) on forskolin-stimulated cAMP accumulation was not modified (37% ± 11% inhibition in the absence and 40% ± 13% inhibition in the presence of CPA; *P* \> 0.2, paired Student\'s *t*-test; [Figure 3(c)](#fig3){ref-type="fig"}), suggesting absence of cross desensitization of cannabinoid CB~1~ receptors by adenosine A~1~ receptors.
Contrasting with CPA, the WIN55212-2 (30 *μ*M) inhibitory effect on forskolin-stimulated cAMP remained virtually unchanged even when slices were preincubated with WIN55212-2 for up to six hours (31% ± 6% inhibition caused by WIN55212-2 for 45 min preincubation and 30% ± 5% inhibition for 6 h preincubation with WIN55212-2; *P* \> 0.05, paired Student\'s *t*-test). Longer incubation periods were not used to avoid losing slice integrity.
4. Discussion {#sec4}
=============
The results obtained in the present work showed for the first time that the inhibitory effect of acute or subchronic coactivation of adenosine A~1~ and cannabinoid CB~1~ receptors on cAMP accumulation is additive in the hippocampus. The results further indicate that the additive inhibitory effects of these receptors are not restricted to excitatory synaptic transmission in the CA1 area \[[@B12]\] but also apply to cAMP formation in the hippocampus. Although a rapid desensitization of the inhibitory action of adenosine A~1~ receptors on cAMP accumulation was observed, this desensitization did not modify the cannabinoid CB~1~ receptor effect on cAMP accumulation.
4.1. Potency and Specificity of A~1~ and CB~1~ Agonists as Inhibitors of cAMP Accumulation {#sec4.1}
------------------------------------------------------------------------------------------
The EC~50~ obtained in the present work in the rat hippocampus (36 nM) for the inhibitory effect of the A~1~ receptor selective agonist CPA, when applied 15 min before forskolin, on cAMP accumulation was similar to that obtained in guinea-pig cerebral cortex (22 nM, \[[@B28]\]). The CB~1~ receptor agonist WIN55212-2 potency for inhibition of forskolin-stimulated cAMP accumulation, obtained in the present work (EC~50~ of 6.6 *μ*M), was also similar to that reported for rat*globus pallidus* slices (EC~50~ between 3 and 10 *μ*M, \[[@B29]\]) and slightly higher than that found in mouse cerebellar membranes (EC~50~ of 1.4 *μ*M, \[[@B23]\]). In hippocampal membranes of guinea-pig the effect of WIN55212-2 (7% maximal inhibition, \[[@B30]\]) was very small to calculate the EC~50~. The inhibitory effect of the A~1~ receptor agonist CPA on forskolin-stimulated cAMP accumulation was prevented by the A~1~ receptor selective antagonist DPCPX, indicating that the effect of the agonist was specific for the adenosine A~1~ receptor. The inhibitory effect of WIN55212-2 on forskolin-stimulated cAMP accumulation was strongly attenuated by the CB~1~ receptor selective antagonist AM251, indicating that the WIN55212-2 effect on cAMP accumulation is mainly mediated by cannabinoid CB~1~ receptor. However, even in the presence of AM251, WIN55212-2 produced a small inhibitory effect on cAMP accumulation. This WIN55212-2 residual effect could be due to (i) activation of cannabinoid CB~2~ receptor; (ii) activation of non-CB~1~, non-CB~2~ receptors. Hypothesis (i) seems unlikely since, although WIN55212-2 is not selective for cannabinoid CB~1~ receptor, CB~2~ receptor is mostly found in peripheral tissues. However, hypothesis (ii) cannot be discarded. In fact non-CB~1~, non-CB~2~ activity of WIN55212-2 has been reported in the hippocampus \[[@B22], [@B31]\], suggesting activation of an unknown receptor.
4.2. Combined Actions of A~1~ and CB~1~ Receptors {#sec4.2}
-------------------------------------------------
We quantified cAMP accumulation to determine how adenosine A~1~ and cannabinoid CB~1~ receptors, when coactivated, modulate adenylyl cyclase activity. We found that when both receptors are simultaneously operating, they exert additive inhibition of adenylyl cyclase activity, which implies that the transduction pathways operated by both receptors do not compete or interfere with each other. If both receptors competed for the same limiting pool of adenylyl cyclase, the combined effect of A~1~ and CB~1~ agonists would be less than additive. These findings agree with previous observations obtained concerning hippocampal excitatory synaptic transmission in the rat (\[[@B12]\], but see \[[@B32]\]), in rat hippocampal membranes (\[[@B33]\], but see \[[@B20]\]) and in whole brain membranes of the mouse where coapplication of A~1~ and CB~1~ receptors agonists additively stimulated \[^35^S\]GTP*γ*S binding \[[@B22]\]. These reports, together with the results obtained in the present study, support an additive effect, in hippocampus, at three different cellular levels when A~1~ and CB~1~ receptors are acutely costimulated: G-proteins, adenylyl cyclase, and excitatory synaptic transmission. In mouse cerebellar membranes, both \[^35^S\]GTP*γ*S binding and inhibition of forskolin-stimulated cAMP accumulation by combined application of A~1~ and CB~1~ receptors agonists were only partially additive, but still the combined effect was greater than the maximal individual effects \[[@B23]\]. In one study A~1~ receptors attenuate CB~1~ receptor-mediated inhibition of K^+^-induced GABA and glutamate release from rat hippocampal synaptosomes \[[@B20]\], which contrasts with the mutually independent inhibitory action of A~1~ and CB~1~ receptors on hippocampal excitatory synaptic transmission found in brain slices \[[@B12]\], probably because availability of signaling molecules in synaptosomes, shared by both receptors, is lower than in brain slices \[[@B34]\]. In C57BL/6J mice (which have high levels of endogenous adenosine) sustained tonic activation of A~1~ receptors prevented CB~1~-mediated inhibition of excitatory synaptic transmission, but not in the rat \[[@B32]\], suggesting differences between species.
Since adenosine A~1~ and cannabinoid CB~1~ receptors mostly couple to identical G*α* ~i/o~ subunits \[[@B35]\] and are both expressed at pyramidal glutamatergic neurons in the hippocampus \[[@B36], [@B37]\], it is not surprising that receptor interference could occur. In fact, A~1~ receptors have less than additive response when interacting with other G~i/o~-coupled receptors, such as group II metabotropic glutamate receptors \[[@B38]\], *α* ~2~-adrenergic receptors \[[@B39]\], and neuropeptide Y receptors \[[@B40]\] in the hippocampus, while, in superior cervical ganglia, the expression of human CB~1~ cannabinoid receptors can sequester G~i/o~ proteins from a common pool and make them unavailable to other G~i/o~-coupled receptors \[[@B41]\]. In rat striatal slices a cannabinoid analogue produced less than additive inhibition of cAMP formation when coapplied with opioid or dopamine D~2~ receptors agonists \[[@B42]\]. On the other hand, additive actions between adenosine A~1~ and *μ*-opioid or GABA~B~ receptor agonists have been described for receptor-mediated G~i/o~ protein activation in hippocampal membranes \[[@B33]\]. Therefore, the additive inhibitory effects of A~1~ and CB~1~ receptors on adenylyl cyclase activity, observed in the present work, suggest that availability not only of G~i/o~ proteins \[[@B33]\], but also of adenylyl cyclase, shared by both receptors, might not be limiting in the rat hippocampus. Another possibility is that compartmentalization of A~1~ and CB~1~ receptors within cells might occur. The scaffold proteins A-kinase anchoring proteins (AKAPs) \[[@B43], [@B44]\], and the lipid raft caveolae \[[@B45]\], have been identified in the hippocampus, where they improve the spatial precision of cAMP-related activity \[[@B46]\].
Formation of heteromers between adenosine A~2A~ and A~1~ receptors has been reported, which explained the interaction between these two receptors \[[@B47]\]. However, the additive and therefore independent action of A~1~ and CB~1~ receptors at the hippocampus observed in the present work does not suggest formation of heteromers between these two receptors.
4.3. Desensitization of A~1~ Receptors {#sec4.3}
--------------------------------------
When applied 95 min before forskolin, CPA failed to modify forskolin-stimulated cAMP accumulation. Therefore, 95 min is a sufficient time period to induce subchronic A~1~ receptor homologous desensitization. In fact, rapid (\<90 min) homologous desensitization of the A~1~ receptor-mediated inhibition of excitatory neurotransmission, induced by hypoxia, has been reported in the rat hippocampus \[[@B48]\]. In smooth muscle DDT~1~ MF-2 cells, uncoupling of A~1~ receptors from G proteins (measured by a decrease in agonist binding) was observed after 30 min exposure to agonist, an effect involving receptor phosphorylation and arrestin binding \[[@B24]\]. In the same cells, desensitization of the A~1~ receptor-mediated inhibition of forskolin-stimulated adenylyl cyclase activity by preincubation with an adenosine A~1~ receptor agonist takes several hours to occur \[[@B49]\].
Sousa et al. \[[@B20]\] reported unidirectional attenuation by A~1~ receptors of CB~1~ receptor-mediated inhibition of glutamate release from hippocampal synaptosomes, while CB~1~ receptors did not affect the A~1~-mediated effect \[[@B20]\], but in this study CPA was present in the incubation medium before WIN55212-2 for over 30 minutes, which may have been enough to trigger desensitization of A~1~ receptors \[[@B24]\]. To evaluate if this apparent unidirectional action of A~1~ receptors on the CB~1~ receptor-mediated effect could be a consequence of heterologous desensitization by the A~1~ receptors, we studied the influence of the CPA incubation period on the WIN55212-2 inhibitory effect. Addition of CPA either 15 min before forskolin (acute stimulation) or 95 min before forskolin (enough to induce subchronic homologous desensitization) did not modify the inhibitory effect of WIN55212-2 on forskolin-stimulated cAMP accumulation, therefore excluding heterologous desensitization of CB~1~ receptors by acute or subchronic adenosine A~1~ receptor activation, at least at the level of cAMP production. However it does not preclude the hypothesis of heterologous desensitization of the receptor response by longer treatment with receptor agonists \[[@B23], [@B50]\], where other downstream effectors may be influenced.
Contrasting with adenosine A~1~ receptors, preincubation with WIN55212-2 for up to 6 h did not induce desensitization of the cannabinoid CB~1~ receptor-mediated inhibition of cAMP production. A previous study in cultured hippocampal neurons indicates that an 18 to 24 h exposure to WIN55212-2 was necessary to produce a significant desensitization of the CB~1~ receptor-mediated inhibition of neurotransmission \[[@B50]\].
4.4. Conclusion {#sec4.4}
---------------
The results obtained in the present work indicate an additive inhibition of cAMP accumulation by adenosine A~1~ and cannabinoid CB~1~ receptors in the rat hippocampus. Furthermore, the effect of CB~1~ was not affected by subchronic A~1~ receptor desensitization. Therefore, the results suggest that receptor cross talk between adenosine A~1~ and cannabinoid CB~1~ receptors does not play a role on acute inhibitory actions of A~1~ and CB~1~ receptors on cAMP production at the rat hippocampus. Since cAMP plays a central role in regulating multiple brain cell functions, it is likely that other additive actions of adenosine A~1~ and cannabinoid CB~1~ receptors, besides inhibition of glutamatergic neurotransmission, might occur at the hippocampus, where a promising cumulative neuroprotective action against neurotoxic insults may occur, which deserves future investigation.
This work was supported by a project grant from the Portuguese Foundation for Science and Technology (FCT, POCI/SAU-FCF/57973/2004). A. Serpa received a scholarship from FCT (SFRH/BD/65112/2009). The authors thank Professor Lori A. Wetmore (William Jewell College) for English editing of the paper.
Conflict of Interests
=====================
The authors declare that there is no conflict of interests regarding the publication of this paper.
{#fig1}
{#fig2}
{#fig3}
[^1]: Academic Editor: George Perry
| {
"pile_set_name": "PubMed Central"
} |
LETTERS TO EDITORS
==================
We found the recent original article published in the Korean Journal of Pain by Seung Yeup Han et al., entitled \"The effect of low-dose ketamine on post-caesarean delivery analgesia after spinal anesthesia\" \[[@B1]\]. The authors expertly detailed an interesting investigation about the prophylactic administration of low-dose intravenous ketamine for reducing postoperative pain and postoperative opioid requirement in patients receiving intravenous fentanyl with patient-controlled analgesia (PCA) after caesarean section. They concluded that \"Intraoperative low-dose ketamine did not have a preemptive analgesic effect and was not effective as an adjuvant to reduce postoperative pain score or opioid requirement in parturients receiving intravenous fentanyl with PCA following caesarean section.\" While this topic is a hot issue in anesthesia and pain medicine, it seems there are some concerns in the methodology of this study that undermine the ability of the reported data to lead to a definite conclusion. Here we note some comments on different points:
\(1\) Accurate estimation of the number of cases is one of the most important points in trial studies. The number of cases in this study was not estimated correctly, and the authors of the mentioned study should estimate the \"power\" of their study in order to compensate for any refusal of data. We can concede that if differences between their study groups (Ketamine group and controls) were not significant during follow-ups, it could be because of the small sample size.
\(2\) In clinical studies, randomization is extremely conclusive and crucial. The authors in this case did not report the method of their randomization (e.g., computer-based table of randomization or other). Therefore, it is not clear whether the two study groups in their study were randomized homologically or not.
\(3\) The authors followed their patients at 2, 6, 24, and 48 h after surgery. Therefore, their study is a prospective clinical trial, but it suffers from the lack of a participation flow diagram. There is no accompanying CONSORT (Consolidated Standards of Reporting Trials) \[[@B2],[@B3]\] fellow chart to provide details on how many parturients declined consent as well as on the overall progress of the participants through the study. The authors affirmed that they enrolled a total of 40 parturients of American Society of Anesthesiologists (ASA) classes I and II undergoing cesarean section in their study and they excluded those parturients who had difficulties communicating, psychological disease, inflammation in the spinal puncture area, etc. However, in the analysis of data, the authors declared that one parturient in the experimental group and three in the control group were excluded due to a switch to general anesthesia or failed spinal puncture. This raises two questions: first, why the exclusion criteria were noted differently in the results section compared to the notes in the methods section, and second, how many patients were lost to follow-up during the study. These points should be stated with more clarity.
\(4\) On the other hand, the exclusion criteria seem to be inadequate for this topic. The authors did not report whether or not they excluded women with a history of malignant hyperthermia \[[@B4]\], hypersensitivity, sensitivity to parabens, and the use of vasoconstrictors which decrease systemic absorption of anesthetic agents \[[@B5]\]. Therefore, the mentioned criteria should be considered by the authors.
\(5\) The authors randomized the patients into the three studied groups and no significant differences in terms of age were reported. However, race, parity, comorbidities, occupation, age of menopause, menopausal hormone therapy, OCP use \[[@B6]\], etc. should be distributed equally among the three groups before any conclusions are drawn. Crucially, there is no mention of these criteria being recorded, nor are they reported in the results section. Thus one cannot be reassured that the studied groups are balanced for these crucial characteristics.
\(6\) In our opinion, the adverse effects of ketamine such as anxiety, bradycardia, chills, edema, drowsiness, hypotension \[[@B7]\], etc. need to be considered by the authors. Generally, the use of ketamine increases the adverse effects which were mentioned previously, and it is a very crucial consideration if there is any possibility of deducing these effects. Additional research is needed in this field.
In conclusion, the mentioned study certainly added evidence to the present literature, and revealed that intraoperative low-dose ketamine did not have a preemptive analgesic effect. Obviously, large-scale clinical trials employing accurate and powerful methods according to the CONSORT Statement and a homogeneous sample with well characterized controls and cases that increase the sensitivity of detecting the associations are required for this topic.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
In the last two decades there have been progresses in understanding thyroid carcinoma histogenesis and biology. Papillary thyroid carcinomas (PTC) represents 80% of thyroid malignancies and is characterized by slow growth and an excellent prognosis \[[@B1]\]. Patients with PTC have the best survival rate of all types of thyroid cancer, with a ten-year rate to 95% \[[@B2]\], or by other studies 5-year survival is 96% and in 10 years 93 % \[[@B3]\]. Differentiated follicular thyroid carcinoma (FTC) constitute 10-17% of malignant thyroid pathology and also have a good long-term prognosis \[[@B4]\]. However, some cases have relatively early relapses, invasion of adjacent structures, lymph node metastases or distant metastases \[[@B5]\].
Thyroid carcinoma that invade local structures are associated with poor prognosis, but the invasion mechanisms are incompletely defined, limiting the development of new therapies. Understanding the molecular mechanisms involved in the progression of thyroid cancer may provide targets for more effective treatment in aggressive thyroid cancer.
Twist is considered to have oncogenic potential by promoting the proliferation and inhibition of apoptosis \[[@B6]\]. It acts in cooperation with Bmi1 inhibiting the expression of epithelial markers such as E-cadherin, and is associated with poor prognosis in a variety of cancers of the head and neck \[[@B7]\]. Loss of E-cadherin expression, in turn leads to loss of cell adhesion and polarity and acquisition of migratory and invasive phenotype of epithelial cells \[[@B8]\].
In this study we aimed the analysis of Twist expression in differentiated thyroid carcinomas and the correlation with clinical and morphological parameters.
Matherial and Method
====================
The study included a total of 43 differentiated thyroid carcinomas selected in a period of nine years from patients hospitalized and operated in Surgery Clinics of Emergency County Hospital of Craiova. he surgical pieces were fixed in 10% buffered formalin, processed by the usual technique with paraffin embedding and Hematoxylin-Eosin stain. Classification of the tumors was made in accordance with literature data \[[@B1]\].
Subsequently the specimens were processed by immunohistochemistry, using citrate antigen recovery buffer pH 6 and a peroxidase-based polymer secondary detection system (Histofine polymer-HRP, Nichirei, Japan, ready to use, code 414151F). The visualization of reactions was done with DAB chromogen (code 3467, Dako), and to validate the results were used positive (tonsil) and negative (omitting the primary antibody) external controls. The tissues analyzed were incubated overnight at 4° C with monoclonal mouse antibody antihuman Twist, diluted 1/2000 (LSBio, code LS-C191858)
We quantified semiquantitatively the expression of Twist by a scoring system which has been assigned two independent specialists on the basis of staining intensity and percentage of positive cells \[[@B9]\]. The intensity score was noted with 0 (no staining), 1 (low intensity), 2 (moderate intensity), and 3 (high intensity). The score of the percentage stained cells was noted with 0 (\<5% positive cells), 1 (6-25% positive cells), 2 (26-50% positive cells), 3 (51-75% positive cells) , and 4 (\> 75% positive cells). The multiplication of the intensity and percentage scores allowed the calculation of the composite final score: 0 (negative), + (1-4) + (5-8), and + + + (9-12). For the statistical analysis, a final staining score of negative or + was considered with low expression and a score of final staining + + or + + + correspond to the group with high expression. The statistical analysis was performed using automatically SPSS10 software and chi-square and Fisher\'s exact tests.
Results
=======
Analysis of clinicopathological parameters of the 43 thyroid differentiated carcinomas indicated an upward trend since the third decade of life, the maximum incidence being after age 60 years, predominantly in females (Table [1](#T1){ref-type="table"}).
Histopathological analysis revealed that the most investigated PTC correspond to conventional type in 23 cases (59%), followed by follicular variant in 13 cases (33.3%), and tall cell variant in 3 cases (7.7%), while for FTC the conventional type was the most common, respectively in 3 cases (75%). The classification of lesions according to the TNM staging indicated 18 cases in stage I, 14 cases in stage II, 7 cases in stage III and 4 cases in stage IVA (Table [1](#T1){ref-type="table"}).
######
The clinicopathological analyzed parameters
----------------------- ------------------ ---------------- ----------------------- ------------------ ------------------------ ---
Tumor type/ Parameter Conventional PTC Follicular PTC Tall cell variant PTC Conventional FTC Clear cell variant FTC
Age \<45 years 10 4 1 1 0
\>45 years 13 9 2 2 1
Gender female 20 12 3 3 1
male 3 1 0 0 0
Stage I 11 5 0 2 0
II 7 5 0 1 1
III 4 2 1 0 0
IVA 1 1 2 0 0
----------------------- ------------------ ---------------- ----------------------- ------------------ ------------------------ ---
In 10 cases of PTC, we found the presence of lymph node metastases, from which 5 cases conventional type, 3 cases of follicular variant and 2 cases tall cell variant.
In the study the analysis of TWIST expression revealed positivity in 34 of the cases, the signal being cytoplasmic and nuclear (Table [2](#T2){ref-type="table"}). Nuclear or nuclear / cytoplasmic immunostain localization was present in 79% of cases, and negative in 21% of the cases. Because TWIST is a nuclear transcription factor, for the immunohistochemical analysis we considering only the nuclear signal.
######
The incidence of thyroid carcinomas expression by TWIST staining scores
------------------------ ---------------- ----------------- ----- ---
Histologic type/\ low expression high expression
Immunoreaction
\- \+ ++ +++
Conventional PTC 5 8 6 4
Follicular PTC 3 5 4 1
Tall cell variant PTC 0 0 2 1
Conventional FTC 1 1 1 0
Clear cell variant FTC 0 1 0 0
------------------------ ---------------- ----------------- ----- ---
The evaluation of staining scores indicated for the majority of tumors values above 3. All variants of the PTC examined, except the tall cell variant, included cases with both low and high expression for TWIST (fig.[1](#F1){ref-type="fig"}). To note that the three cases of PTC tall cell variant presented high expression for TWIST.
Thus, in the conventional papillary carcinoma, the average score of positivity was 5.83, with mostly moderate / high intensity of reactions and a percentage of the marked cells of 20-80% (fig.[1](#F1){ref-type="fig"}A-B). Regarding follicular and tall cell variants of PTC, the positivity average score was 5.60, respectively 7.00, intensity mostly moderate / high and a percentage of labeled cells between 20-80% and respectively 50-70 % (fig.[1](#F1){ref-type="fig"}C-D). Comparative, FTC indicated a positivity score of 4.66, the reactions presenting moderate / high intensity, with a percentage of labeled cells of 25-65% (fig.[1](#F1){ref-type="fig"}E-F).
{#F1}
We found no differences in Twist expression depending on the type (p = 0.571, Fisher\'s Exact Test) or subtype of the tumor (p = 0.433, chi square test). Also, we found no statistical association of Twist immunoexpression in relation to gender and age groups of patients.
The high grade Twist expression was associated with cases of thyroid carcinomas that presented invasion of adjacent structures (p=0.002, chi square test, fig.[2](#F2){ref-type="fig"}A).
Also high grade Twist expression was associated with thyroid carcinomas that have associated lymph node metastases (p=0.01, chi square test, fig.[2](#F2){ref-type="fig"}B)
{#F2}
Discussion
==========
Twist is a member of the class of transcription factors, one of the molecules involved in the regulation of epithelial-mesenchymal transition (EMT) via regulation of key proteins that maintain the characteristics of epithelial cells and which that gives a mesenchymal phenotype \[[@B10]\]. Twist is overexpressed in many tumor types, including breast, gastric, hepatocellular, prostate \[[@B11]\], lung \[[@B8]\], esophagus \[[@B12]\] and bladder \[[@B6]\]. In some studies Twist overexpression is associated with high-grade cancer aggressiveness and poor patient survival rate \[[@B13], [@B14]\].
Epithelial-mesenchymal transition (EMT) is a dedifferentiation program that converts epithelial cells into mesenchymal phenotype and is involved in cancer progression \[15\]. More recent studies have indicated that EMT also plays a critical role not only in tumor metastasis, but also in tumor recurrence, processes that are considered to be closely related to the biology of cancer stem-like cells or initiation cancer cells \[[@B14]-[@B18]\]. Twist is a transcription factor that promotes EMT associated with tumor metastasis, cancer stem cells and drug resistance \[[@B19],[@B20]\].
Twist role in tumor progression is associated with the metastatic process \[[@B10]\]. Twist overexpression increases the invasive and metastatic ability of cancer cells by downregulation of E-cadherin expression and induction of EMT \[[@B10]\]. Twist plays an important role in some of the processes involved in the development of metastasis, such as angiogenesis, invasion and extravasation of chromosomal instability, and also protects cancer cells from apoptosis \[[@B21]\].
In our study we analyzed the Twist expression in 43 differentiated thyroid carcinomas. Both the variants of papillary and follicular thyroid carcinomas presented variable immunoreactions, with cases with low and high Twist expression with no statistically significant differences in relation to tumor type or subtype.
In a study conducted in 2013, Ni Wang et al. analyzed 121 papillary thyroid carcinomas and found low Twist expression in 56.6% of the cases, and high expression of the protein in 43.4% of the cases, without difference between tumor subtype, respectively, in conventional, follicular, tall cells and oncocitic variants; in the same study authors indicate non-significant differences of Twist expression in relation with pTNM stage, but a high protein expression in metastasing papillary carcinomas \[[@B9]\]. In our study, low Twist expression was present in 53.8% of PTC and high expression in 46.2% of these lesions, most commonly, but not statistically significant in tall cell variant. Also, 75% of FTC cases present low expression and in 25% of cases we observed a high expression of Twist protein.
Twist expression is differently reported in the literature studies. Thus, D Buehler et al. in a study that investigated the expression of Twist 1 in 27 follicular carcinomas and 28 papillary thyroid carcinomas, including tall cell and follicular variants found no reactions in all cases, the immunostaining being present only in anaplastic forms of tumors \[[@B22]\]. On the contrary, CN He et al. in 2008, on a lot of 50 papillary thyroid carcinomas indicate Twist positivity in all cases, the marker expression being associated tumor aggressiveness \[[@B23]\].
In our study we found differences of Twist expression in relation to tumor stage (stage I / II versus stage III / IV) and the presence of metastasis, the protein expression being significantly higher in invasive tumors and in cases of tumors that had metastasized. In this respect, the literature indicates Twist as a marker of tumor aggressiveness, and associated with forms of invasive and metastatic carcinomas, including thyroid tumors \[[@B9],[@B13],[@B14],[@B22],[@B23]\].
Conclusions
===========
In differentiated thyroid carcinomas, Twist overexpression is associated with an invasive and metastatic immunophenotype. Identifying of high grade Twist expression in various forms of differentiated papillary thyroid carcinomas supports for a personalized therapeutic attitude to increase patient survival.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijerph-15-02367}
===============
It is estimated that around 15% of schoolchildren have a chronic health issue \[[@B1-ijerph-15-02367],[@B2-ijerph-15-02367]\] and that, due to these problems, they can present some degree of inhibition for full integration in their school activities, giving rise to behavioral problems \[[@B3-ijerph-15-02367]\]. Generally, the school population has increasingly more chronic pathologies and more deficiencies in health issues. In addition, there is a progressive increase in the number of foreign students with different pathologies and in situations of economic and social marginalization. For all these reasons, the training of health professionals requires greater specialization and adaptation \[[@B3-ijerph-15-02367]\], taking into account that there is an undocumented need for giving more specialized attention within different educational settings to the problems that arise in paediatric populations with associated chronic pathologies.
In response to these new health demands, having an in-school nurse has been considered as an interesting option for those responsible for both the healthcare and education sectors. The nurse could play the role of health agent, either from the primary care or educational centre position, being a professional with full responsibility and potential in terms of education and health promotion development in the school community, in addition to acting as a reducing factor of stress and anxiety \[[@B4-ijerph-15-02367],[@B5-ijerph-15-02367],[@B6-ijerph-15-02367]\].
The figure of the school nurse has been implemented in some countries such as France, the United Kingdom, Germany, and Switzerland in Europe and has been fully established in the United States for many years \[[@B7-ijerph-15-02367]\]. However, in Spain, a school nurse is only present in some autonomous communities and in ad hoc situations.
For this reason, our aim is to identify vital risk health issues and complications among students as well as the teachers' training on and identification of the same.
2. Materials and Methods {#sec2-ijerph-15-02367}
========================
This is an explanatory sequential design in two stages and our aim is to use qualitative data to help us explain the quantitative results initially obtained so as to better develop the quantitative results \[[@B8-ijerph-15-02367],[@B9-ijerph-15-02367]\]. An explanatory sequential design is a mixed method.
Mixed methods research is increasingly being used as a methodology in the sciences to gain a more complete understanding of issues and hear the voices of participants \[[@B10-ijerph-15-02367],[@B11-ijerph-15-02367]\]. Mixed methods research is the collection and analysis of both qualitative and quantitative data and its integration, drawing on the strengths of both approaches \[[@B12-ijerph-15-02367],[@B13-ijerph-15-02367]\] as a way to represent and facilitate the integration of qualitative and quantitative data in mixed methods studies \[[@B14-ijerph-15-02367]\].
Within these studies, preliminary quantitative results are used to report on the phenomenological approach within the second phase of the research. Therefore, quantitative data either provide guidance towards the phenomena discovered within the preliminary quantitative phase or help identify participants for the phenomenological phase which can provide information on the experiences of respondents \[[@B15-ijerph-15-02367],[@B16-ijerph-15-02367],[@B17-ijerph-15-02367],[@B18-ijerph-15-02367],[@B19-ijerph-15-02367],[@B20-ijerph-15-02367],[@B21-ijerph-15-02367],[@B22-ijerph-15-02367]\].
2.1. Quantitative Phase Methodology {#sec2dot1-ijerph-15-02367}
-----------------------------------
### 2.1.1. Design and Scope of the Study {#sec2dot1dot1-ijerph-15-02367}
This was an observational, descriptive, and cross-sectional study on an intentional nonprobabilistic sample of 3246 teachers both from general education centres (preschool, primary, secondary, high school diploma, vocational training, special education, and distance learning) and special regime systems (school of plastic arts and design, school of music and dance, official language schools, drama schools, and sports schools) in Spain. The recruitment of participants was conducted between January and March 2017. Only teachers over 65 were excluded.
### 2.1.2. Estimation of Sample Size {#sec2dot1dot2-ijerph-15-02367}
The set of teachers registered in the Ministry of Education of Spain was taken into account for the calculation of the sample size, with a total of 695,598 teachers (EducaBase) in the school year 2016--2017. Fifty percent were used as the reference prevalence, as it was a multiple choice questionnaire and this was the most demanding prevalence. The confidence level was 95%, with an absolute error of 2%, resulting in a minimum of 2393 participants.
### 2.1.3. Sources of Information {#sec2dot1dot3-ijerph-15-02367}
An ad hoc data collection notebook was designed for this purpose. It included an explanation of the aims and purposes of the study and ensured anonymity, confidentiality, and further ethical guarantees. This questionnaire ([Appendix A](#app1-ijerph-15-02367){ref-type="app"}) was sent to various associations or groups of teachers through social networks as well as to the various provincial delegations of education which, in turn, referred this questionnaire to all educational centres, and the contribution of the regional teaching institution was especially important for the task of diffusion distribution of the questionnaire to all centres and responsible teachers in a systematic manner.
### 2.1.4. Study Variables {#sec2dot1dot4-ijerph-15-02367}
The questionnaire collected information on sociodemographic variables about the professional profile of teachers, their experience in students' health risk situations, the existence of children with health problems in the current academic year, the degree of training for and concern over these situations, as well as the relevance of the hypothetical figure of the school nurse.
### 2.1.5. Statistical Analysis {#sec2dot1dot5-ijerph-15-02367}
First, descriptive statistical analysis using absolute and relative frequencies for the qualitative variables was conducted.
For the bivariate analysis, Pearson's chi-squared test was used when both variables were qualitative, and the Mann-Whitney U nonparametric test was used when studying the relationship between the variables with ordinal response and qualitative dichotomous ones.
All analyses were performed with the statistical package SPSS 24.0 (SPSS Inc., Chicago, IL, USA).
### 2.1.6. Ethical Aspects {#sec2dot1dot6-ijerph-15-02367}
This study was designed in accordance with the Helsinki Declaration adopted by the World Medical Association (WMA). The completion of the questionnaire was completely voluntary and anonymous.
Approval by the Ethics in Research Committee Mancha Centro Hospital with number C-149 February 2017 was granted before starting the research.
2.2. Qualitative Phase Methodology {#sec2dot2-ijerph-15-02367}
----------------------------------
A descriptive, qualitative, phenomenological design based on Giorgi's method \[[@B23-ijerph-15-02367]\] was developed. It was aimed at teachers and its information analysis was carried out using the phenomenological method of Amedeo Giorgi \[[@B24-ijerph-15-02367]\].
This was a purposive sampling \[[@B25-ijerph-15-02367]\], conducting interviews with those teachers who were in contact with the research team after having seen the appeals made in social networks during the months of January--March 2017, and although the theoretical sample size was insufficient to guarantee external validity in terms of other models of research \[[@B26-ijerph-15-02367]\], it was sufficient to complete all categories, including participants from different sociodemographic characteristics.
The collection of data through interviews was recorded on audio files with a password, and only the authors had access to the same \[[@B27-ijerph-15-02367]\].
The only criterion for inclusion was being a teacher with an average age of 21--65 years.
A total of 16 semistructured interviews were conducted, collecting a number of sociodemographic characteristics such as age or sex following a semistructured script we created from the quantitative phase to modulate the interviews, with a duration of between 40 and 60 min, which were transcribed in their entirety. The participants did so voluntarily and with informed consent. The teachers were identified by codes to respect their anonymity, identifying discourses with the letter E (for "interview" in its Spanish form entrevista), followed by a sequential number from 1 to 16.
All interviews began with an open question to invite participants to narrate their experiences as teachers and the need for a nurse in their centres so that they could focus on the phenomenon in question \[[@B23-ijerph-15-02367]\]: Could you tell me what it is for you to be a teacher and if you think it is necessary to have a nurse at educational centres? Participants were encouraged to freely narrate their experience, and the interviewer would freely follow his script to encourage them.
The credibility, auditability, and transferability methodological rigour criteria \[[@B28-ijerph-15-02367]\] were taken into account during the process.
3. Results {#sec3-ijerph-15-02367}
==========
3.1. Quantitative Phase Results {#sec3dot1-ijerph-15-02367}
-------------------------------
This work involved a total of 3246 teachers in Spain, of which 65.8% (2136) were women, and the most numerous age group consisted of teachers aged between 31 and 40 years (34.8% (1131)). 74.3% (2409) had been teaching for more than 10 years. Further information can be found in [Table 1](#ijerph-15-02367-t001){ref-type="table"}.
As to whether teachers had received specific training in urgent action on children's health problems, 54.9% (1783) had never participated in such training, 13.2% (429) had received training in the last 5--10 years, 21.0% (683) in the last 1--5 years, and 10.8% (351) in the last year.
In this sense, the relationship between the training received and the profile of teachers was analysed, observing a statistically significant association with the age of teachers and the years of practice (*p* \< 0.001). In the light of this, teachers who were older and had more years of practice had a higher percentage of training conducted. However, from 50 years of age and with more than 20 years of teaching, there was a decline in the amount of training. It should also be noted that, regardless of these results, there is a percentage higher than 50% in all categories of teachers who have never received this training ([Table 1](#ijerph-15-02367-t001){ref-type="table"}).
Then, teachers were asked about whether, in their professional career, they had experienced a series of health problems requiring urgent action. To this question, the most frequently observed problems were dizziness and/or syncope with 75.4% (2447), followed by injuries requiring sutures or plaster splints with 61.2% (1986), and hypoglycaemia with 52.2% (1693).
In addition, teachers were asked about their level of concern over the occurrence of any of these health problems, and we found that 56.6% (1837) of teachers were very concerned about it and 33.0% (1071) were fairly concerned ([Table 1](#ijerph-15-02367-t001){ref-type="table"}).
They were also asked to what extent they were prepared to face any of these problems, of which 16.4% (533) answered they were not prepared and 46.9% (1523) expressed feeling scarcely prepared ([Table 1](#ijerph-15-02367-t001){ref-type="table"}).
In this case, relating the degree of concern and the perception of their preparedness with the completion of training in urgent action, we found a statistically significant association. Teachers who had received training had increased their perception of preparedness (*p* \< 0.001) and concern (*p* = 0.027) with respect to those not trained.
Then, so as to have an overall view of the health state of children and adolescents attending educational centres in Spain, teachers were asked if there was at least one child with a series of health issues in their classrooms. Results can be found in [Table 2](#ijerph-15-02367-t002){ref-type="table"}, which highlight that 69.6% (2258) of teachers had students with attention-deficit hyperactivity disorder in their classrooms, 62.8% (2038) with food allergies, and 56.8% (1845) with asthma.
Finally, when asked to what extent could the presence of a specialized nurse be relevant in educational centres to deal with these situations, 51.9% (1684) stated that it would be very important and 29.9% (972) quite relevant ([Table 1](#ijerph-15-02367-t001){ref-type="table"}). In this sense, the trained teachers give greater relevance to the figure of the school nurse than those who are not trained (*p* = 0.001).
3.2. Qualitative Phase Results {#sec3dot2-ijerph-15-02367}
------------------------------
### 3.2.1. Participants {#sec3dot2dot1-ijerph-15-02367}
The analysis of the interviews allowed us to stablish four categories that are representative of the life experience of teachers and the need to have a school nurse in educational centres. These categories have been named as: Need when facing reality, Fears, Responsibility, and Limitations as a teacher ([Table 3](#ijerph-15-02367-t003){ref-type="table"}).
### 3.2.2. Need When Facing Reality {#sec3dot2dot2-ijerph-15-02367}
Our study teachers felt the need for a nurse in educational centres, but at the same time, there was a feeling of disappointment in terms of their reality regarding the need for teachers. Therefore, on the one hand, they believe it necessary to have a nurse in their centres, but surely not at the expense of reducing their own staff or hiring fewer teachers in order to have a nurse in each centre. There is a clear dilemma and the need for an unfounded choice between having a school nurse and hiring more teachers:
> "*It would be great, but it would be better to have the necessary number of teachers.*"
>
> (E3)
> "*I believe that a nurse by centre should be mandatory, but more teachers would also be good.*"
>
> (E5)
> "*It would be great! But in the light of how things are usually done, they could even hire a nurse by centre and tell them to be substitute teachers.*"
>
> (E9)
### 3.2.3. Fears {#sec3dot2dot3-ijerph-15-02367}
From the interviews, we obtained data regarding being a teacher in charge of taking caring of children with chronic diseases; this fact leads them to have a number of fears based on the lack of knowledge about what to do in certain cases which can occur throughout their professional career:
> "*To be honest, a nurse is quite necessary in our schools and high schools. I have found myself, more than once, due to my specialty, in the situation of students with deep cuts, and it feels really bad. You don't really know what to do, and with serious mishaps we have to perform a quick cure without any means (I teach cooking and pastry, and in many centres, there is no first-aid kit), and we need to go to the hospital in a rush\...Kids with epilepsy, diabetes, cardiac problems.*"
>
> (E6)
> "*I have a student with epilepsy and am always afraid of something bad happening. Even with indications of how to intervene, I don't know if I would be able to do it properly. I am completely in favour of a nurse in every school.*"
>
> (E8)
### 3.2.4. Responsibility {#sec3dot2dot4-ijerph-15-02367}
In the data reduction, the term "responsibility" recurrently emerged. Teachers believe that assisting a student is not their responsibility, since it implies burdens they do not want to carry and are reluctant to do so; they expressed the need of a school nurse, although their words carry a number of nuances, such as considering the omission of assistance a crime.
> "*I would also like it if there were a nurse in every educational centre, and even more after knowing that, in case of emergency, we will have to inject what they need (if their life is in danger) and if we don't, it is a crime.*"
>
> (E7)
> "*It is crime, and it can also be the ruin of you if the judge believes that you did not act as the parents would have done. Not to mention if you rush to seek help\...They can tell you you've abandoned the student.*"
>
> (E5)
As an emerging expectation in all the speeches, the concerns about the limitations they have in relation to their work and way of acting arise. Teachers see the need for training in first-aid courses, although they continue to think there is the need for a school nurse:
> "*Well, so they should then require a first-aid course as a minimum. Rather than scoring online courses taken just because, they should score first aid. I sincerely believe this is a limitation.*"
>
> (E11)
> "*We have taken the trouble to take first-aid courses, but it would be better to have a qualified person to do so.*"
>
> (E15)
4. Discussion {#sec4-ijerph-15-02367}
=============
In the present study, the opinions of teachers regarding their perception of the degree of training and concern in first aid with children was evaluated, as well as the need for a nurse in educational centres of Spain.
Regarding this, it is highlighted that more than half of the teachers have not received any training, with percentages of high concern of 56.6% and with a perception of deficient training to deal with these problems of 46.9%. Hence, the presence of a nurse at educational centres was very positively valued. Based on these results during the qualitative process, four broad categories have been set: need when facing reality, fears, responsibilities, and limitations as a teacher.
On the other hand, we can highlight that up to 75.4% of the teachers have witnessed dizziness or syncope, and 61.2% have witnessed trauma in children that have required stitches or plasters, concurring with other studies \[[@B29-ijerph-15-02367],[@B30-ijerph-15-02367]\]. We must here add that 69.6% of teachers declare having a student with chronic problems. This situation creates anxiety for some teachers, as is highlighted in the qualitative results that coincide with other authors such as Gomez \[[@B31-ijerph-15-02367]\] and Fernandez \[[@B32-ijerph-15-02367]\]. These refer to teachers who acknowledge having or having had a student with a chronic pathology, and this issue causes them some degree of uncertainty.
These situations have as a consequence that the incorporation of nurses in schools is not only perceived as relevant by teachers, but that, as other authors such as Calvo \[[@B33-ijerph-15-02367]\] and Díez \[[@B34-ijerph-15-02367]\] have found, between 72% and 86% of parents, respectively, claim that the presence of a nurse at the educational centre is necessary.
The limitations of our study include those arising from the absence of a nonrandom sampling procedure. The attitudes of the people who answered voluntarily could have been more favourable than that of those who did not participate. It is possible that the participation in this study reflects a greater involvement of teachers in need of a school nurse or just the opposite. Despite this, we believe that it is a quite accurate portrait of reality because the qualitative part of the study confirms that teachers feel there is a lack of knowledge and training in health education, although some of them allege that nurses should be the ones having training in schools, and that the responsibility should not fall on them as teachers.
Implications for Nursing and Health Policy
------------------------------------------
Nevertheless, an alternative to the school nurse could be developed through an ambitious training and sensitization plan for teachers. However, these plans could not guarantee that all teachers are trained, as can be seen in our results, and, on the other hand, it is difficult to achieve the appropriate capacity for technical aspects such as the preparation of medication and its administration \[[@B35-ijerph-15-02367]\]. In addition, the school nurse would not only be limited to resolving specific mishaps, but they would have a more ambitious role in the field of promotion and prevention. Duties would range from a closer monitoring programme of healthy children to teaching in seminars both for teachers and students, and even the application of cures or specific techniques at schools that would prevent school absenteeism due to the transfer to and from the health centres, etc. This aspect is particularly important when taking into account the results observed in our study on the frequency of children with pathologies. It would be interesting to develop a pilot project to assess the impact that the presence of school nurses at all levels has on: public health, occupational risk prevention, reducing hospital admissions, school absenteeism, prevention of risk behaviours, occupational quality of teachers, anxiety in teachers and parents, and even, in economic terms, a training plan for teachers in this subject or any other alternative.
To incorporate a nurse into Spanish schools at present, it would only be necessary for the political leaders in the management of the educational centres to be convinced of the importance of the creation of this figure. In Spain, there are nursing professionals with specific degrees in community nursing and paediatric nursing who could assume this role. On the other hand, the creation of this new figure could result in savings for the health system, as other studies have shown \[[@B36-ijerph-15-02367]\]. The implementation of the teaching nurse should be progressive. Initially, pilot experiences should be developed in different centres before their systematic implementation to establish the functions of the teaching nurse and the evaluation of their economic impact.
5. Conclusions {#sec5-ijerph-15-02367}
==============
As a conclusion, we find that there is a significant percentage of children with diseases that often require specific care and that there is a high probability that teachers, throughout their professional lives, have to deal with situations of vital urgency. In this sense, teachers in Spain are not trained in urgent action. In addition, they have the perception of having low training and this is an important concern. The presence of professionals in educational centres seems a relevant option, although they manifest a dilemma regarding the fear that these professionals may imply a decrease of teaching professionals.
Conceptualization, A.H.-M. and J.R.-A.; Methodology, J.G.-S.; Formal Analysis, A.H.-M. and N.I.-T.; Writing---Original Draft Preparation, M.D.-M.; Writing---Review & Editing, J.R.-A.; Supervision, G.A.-A.; Project Administration, J.G.-S.
This research received no external funding
The authors declare no conflict of interest.
1. Age (years):
2. Sex ManWoman
3. How many years have you been practicing as a teacher (approximately)?
4. Course to which he teaches his classes at present:
5. Have you participated in any specific training on urgent action in the face of a health problem in children? NoYes
6. If yes, when was the last time? In the last yearBetween 1 and 5 yearsBetween 5 and 10 years
7. Mark with a cross if you have witnessed any of these problems in the children in your class. Cardiorespiratory arrestAllergic reactionHypoglycaemia (low blood sugar)Epileptic seizuresTrauma that requires sutureDizziness or syncopeChoking
8. To what extent do you worry that any of these health problems may occur in your presence? NothingLittle bitSomethingQuiteMuch
9. To what extent are you prepared to face any of these health problems? NothingLittle bitSomethingQuiteMuch
10. Mark with a cross if you currently have a child in your class with these health problems ObesityAtopic dermatitisHeart problemsDiabetesCeliac disease (gluten intolerance)EpilepticFood AllergyAsthmaTDHMetabolic disease (hypothyroidism, phenylketonuria, etc.)Neurological disease (For example: cerebral palsy)Other psychiatric problems (autism, depression, etc.)
11. To what extent do you consider that the presence of a specialized nurse professional could be relevant to deal with the emergency situations and health problems indicated above? NothingLittle bitSomethingQuiteMuch
ijerph-15-02367-t001_Table 1
######
Relation between teacher profile and training received.
Characteristic Total Training Received *p*-Value
----------------------- ------------- ------------------- ------------ -------------
Sex 0.051 ^a^
Male 1110 (34.2) 636 (57.3) 474 (42.7)
Female 2136 (65.8) 1147 (53.7) 989 (46.3)
Age 0.001 ^a^
\<30 205 (6.3) 137 (66.8) 68 (33.2)
31--40 1131 (34.8) 606 (53.6) 525 (46.4)
41--50 1121 (34.5) 592 (52.8) 529 (47.2)
\>50 789 (24.3) 448 (56.8) 341(43.2)
Years as a teacher \<0.001 ^a^
\<1 year 96 (3.0) 73 (76.0) 23 (24.0)
1--5 years 275 (8.5) 184 (66.9) 91 (33.1)
6--10 years 466 (14.4) 263 (56.4) 203 (43.6)
11--15 years 775 (23.9) 405 (52.3) 370 (47.7)
16--20 years 620 (19.1) 306 (49.4) 314 (50.6)
21--30 years 716 (22.1) 395 (55.2) 321 (44.8)
\>30 years 298 (9.2) 157 (52.7) 141 (47.3)
Degree of concern 0.027 ^b^
High 1837 (56.6) 980 (55.0) 857 (58.6)
Quite high 1071 (33.0) 603 (33.8) 437 (32.0)
Medium 244 (7.5) 147 (8.2) 97 (6.6)
Low 85 (2.6) 46 (2.6) 39 (2.7)
No concern 9 (0.3) 7 (0.4) 2 (0.1)
Degree of training \<0.001 ^b^
High 19 (0.6) 6 (0.3) 13 (0.9)
Quite high 151 (4.7) 34 (1.9) 117 (8.0)
Medium 1020 (31.4) 423 (23.7) 597 (40.8)
Low 1523 (46.9) 903 (50.6) 620 (42.4)
No training 533 (16.4) 417 (23.4) 116 (7.9)
Need for school nurse 0.001 ^b^
High 1684 (51.9) 980 (53.3) 857 (46.7)
Quite high 972 (29.9) 603 (56.3) 437 (47.5)
Medium 356 (11.0) 147 (60.2) 97 (39.8)
Low 174 (5.4) 46 (54.1) 39 (45.9)
No need 60 (1.8) 980 (53.3) 857 (46.7)
^a^ Pearson's chi-squared test; ^b^ Mann-Whitney U test.
ijerph-15-02367-t002_Table 2
######
Students health issues in the current course, as notified by teachers.
Issue (*n*/%)
------------------------------------------ -------------
Metabolic disorders 337 (10.4)
Heart problems 658 (20.3)
Epilepsy 671 (20.7)
Serious neurological problems 1051 (32.4)
Diabetes 1076 (33.1)
Atopic dermatitis 1206 (37.2)
Obesity 1382 (42.6)
Psychiatric disorders 1413 (43.5)
Celiac disease 1416 (43.6)
Asthma 1845 (56.8)
Food allergy 2038 (62.8)
Attention-deficit hyperactivity disorder 2258 (69.6)
ijerph-15-02367-t003_Table 3
######
Sociodemographic characteristics of teacher (qualitative phase).
Interview *n* = 16 Age Time Working as a Teacher
-------------------- ----- ---------------------------
1 (Male) 24 1 Year
2 (Female) 22 1 Year
3 (Female) 52 20 Years
4 (Male) 33 5 Years
5 (Male) 47 15 Years
6 (Male) 28 3 Years
7 (Female) 41 11 Years
8 (Female) 49 22 Years
9 (Male) 33 6 Years
10 (Female) 41 14 Years
11 (Female) 25 2 Years
12 (Female) 60 30 Years
13 (Male) 43 7 Years
14 (Male) 62 32 Years
15 (Female) 30 7 Years
16 (Male) 36 5 Years
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec0005}
===============
Peste des Petits Ruminants (PPR) is a viral disease of sheep and goats associated with high morbidity and mortality in affected flocks and found in large parts of Africa and Asia ([@bib0015]). In 2015, FAO and OIE launched a Global Control and Eradication Strategy for PPR which utilises a progressive stepwise approach with four stages: 1) Assessment of epidemiological situation; 2) Implementation of control activities; 3) Eradication; 4) Post-eradication ([@bib0025]). The strategy is based on vaccination, which is a key activity in stages 2 and 3. The PPR vaccine promoted by the programme stimulates longevity of antibodies post vaccination and is believed to confer immunity for at least three years, longer than the lifespan of many small ruminants ([@bib0020]). Despite this immunogenicity, maintaining effective vaccination coverage will be a major challenge due to issues including: a large at risk population; high population turnover; seasonal fluctuations in animal numbers; inaccessibility of herds; high rates of animal movements; lack of co-operation from livestock owners including a reluctance to vaccinate; relatively unknown marketing systems ([@bib0005]). As part of the progressive control, the vaccine supply chain must also be robust and responsive to ensure vaccine availability to eliminate virus circulation. A risk-based approach to vaccination is desirable especially in the early stages of control ([@bib0040]) and is likely to be more cost effective. However, surveillance systems in affected countries are currently unable to provide sufficient understanding of PPR epidemiology in different settings to use this approach globally ([@bib0005]). In addition, vaccination is not synonymous with immunisation. Reasons for an animal not responding effectively to a vaccination may include: poor vaccine quality; inadequate maintenance of the cold chain; and in young animals, maternal immunity interfering in vaccine response. These issues contribute to uncertainties in the estimated cost of global control and eradication, which is made worse by the lack of reliable information on the cost of vaccination per animal in different production systems, with the economic analyses relying on assumptions without field validation. For example, when evaluating the economic impact of global eradication of PPR, Jones et al ([@bib0035]) used estimates based on the cost of rinderpest vaccination because the cost data for PPR vaccination in small ruminants were not available in the literature. Tago et al ([@bib0045]) recently adapted a costing tool used by the World Health Organisation to livestock vaccination using PPR in Senegal as an example. The approach was useful for planning resource allocation although these estimates were not derived from field observations or externally validated.
For the PPR global eradication programme to be successful, decision-making on the allocation of resources and their effectiveness could be informed by collecting data and experiences from actual vaccination programmes. In particular, cost data of implementing vaccination will be vital to identify weaknesses of vaccine distribution and delivery. This is regularly carried out in human vaccination programmes, yet has not been formalised in animal health settings ([@bib0010]). A recent review on the use of economic and social data in veterinary vaccine development revealed the relative paucity of socio-economic analyses in this field ([@bib0050]). Such data are the basis for truly evidence-based benefit-cost analyses or assessing the cost-effectiveness of new technologies ([@bib0030]). Given the important gap in vaccination cost information from animal health in general, and PPR in particular, the aim of this study was to collect data from ongoing PPR vaccination campaigns in Ethiopia to provide information on the cost in different production systems and to estimate the annual cost to the Ethiopian economy. During this process, a data collection methodology was developed that could be applied in other countries and for other livestock diseases.
2. Materials and methods {#sec0010}
========================
2.1. The cost model {#sec0015}
-------------------
The cost of vaccination was assumed to include the following components aggregated using the model:Where:
Vaccine cost = vaccine price + reconstituting saline price
Vaccine transport cost = transport truck cost + fuel cost + transporting personnel cost
Storage cost = (vaccine storage freezer depreciation rate + electric power cost) \* time vaccine spent in storage
Field delivery cost = field vaccination personnel cost + field transport cost (car, fuel and car maintenance cost) + Material cost (price of consumables + depreciation of durable vaccination and cold chain equipment)
Farmers' time = hours spent by the farmers in getting their flocks vaccinated \* wage per hour
Coordination cost = salary, per diem and transport cost of the local vaccination coordinator.
Mobilization cost = cost of personnel doing mobilization of stock owners + other media broad cast costs
Vaccine wastage= (vaccine cost \* missed shots) + (vaccine cost \* vaccine discard)
The different delivery components (i.e. transport, storage, delivery, co-ordination and mobilization) were calculated on a per dose basis with the denominator being the total number of doses used during the data collection period (i.e. including any wastage). The farmer's time was included as an opportunity cost assuming this time could be spent doing manual labour and was also calculated based on the number of farmer's animals vaccinated.
2.2. Data collection {#sec0020}
--------------------
Data for the cost model were collected from vaccination conducted in four regional states in Ethiopia: Afar, Amhara, Somali and Southern Nations Nationalities and Peoples (SNNP) regional states. In each region, data were collected by one of the authors (WJ) from a single district: Ada'ar district from Afar, Erer district from Somali, Dasenech district from SNNP and Burie from Amhara ([Fig. 1](#fig0005){ref-type="fig"}). The choice of region and district was based on availability of active PPR vaccination programs during the study period (February to August 2017).Fig. 1Map of Ethiopia showing the regional boundaries and districts where data were collected from PPR vaccination campaigns to estimate the cost of vaccination.Fig. 1
The vaccinations in Afar, Somali and SNNP regions were within pastoral systems. These were implemented by the Food and Agriculture Organization of the United Nations (FAO) as part of a progressive stepwise approach against PPR in the lowlands of Ethiopia under the EU funded-project "Pursuing Pastoral Resilience through Improved Animal Health Service Delivery in Pastoral Areas of Ethiopia". The vaccination in these regions was organized and implemented in the form of campaigns covering entire pastoral areas delineated by participatory disease search teams around an area where PPR virus was confirmed by pen-side test (Rapidtest). The vaccination in the Amhara region, which has a mixed crop-livestock system, was also reactive but in response to passive surveillance performed by individual districts using the Ethiopian government resources. In this region annual regular preventive vaccination had occasionally been delivered but not during the study period.
A data collection checklist was developed to collect data on the components of the economic cost model and it was applied through a combination of interviews and field observations in each district. Cost data on vaccine transport, vaccination coordination and mobilization were collected from interviews with vaccination coordinators. Cost data on field delivery, vaccine wastage from missed shots, and farmer's opportunity cost from time getting their animals vaccinated were collected by personal observation of the study enumerators (animal health personnel recruited for data collection) during vaccination using a prepared format. Field observations of at least one vaccination team were carried out for three days in each district except in Bure where the observation was for only one day. During field observation, enumerators recorded the number and composition of vaccination personnel in a team and the number of animals vaccinated by the team per day. Additionally, close observations of injections from the beginning to the end of a full vaccination gun (typically 30 doses) allowed recording of the proportion of missed injections. Farmers attending the vaccination were asked the time taken to bring their animals and return to their origin including the waiting time at the vaccination site.
2.3. Data analysis {#sec0025}
------------------
All field data were manually recorded and initially entered into a MS Excel spreadsheet before being transferred into RStudio (v1.0.143) for analysis. The cost for each component of the cost model was calculated and aggregated to estimate the cost per dose of vaccine used. Probability distributions for each cost component were fitted based on the collected data and were applied in the estimate of the PPR vaccination costs at population level stratified by production system (pastoral versus mixed-crop livestock). Livestock population data were available online from the Central Statistical Agency of Ethiopia (<http://www.csa.gov.et/index.php/survey-report/category/348-eth-agss-2016>, Appendix A). The simulation was run 100 times.
Based on observations in the field there was noticeable vaccine discard (i.e. vaccine remaining in syringes and bottles disposed when moving between vaccination sites or at the end of the day) although specific records were not kept. To account for this wastage, it was assumed that for each site visited up to ten bottles were opened and half of one of these bottles was discarded. This indicated an average of 14.6% vaccine discard. This extra cost of vaccine was added to the cost model. Since the number of bottles used, and thus the proportion of wastage is expected to vary with each site (proportion of wastage decreases exponentially as the number of bottles opened increase), this was fitted as an exponential distribution in the final model (Appendix B).
3. Results {#sec0030}
==========
Four vaccination campaigns were observed as part of the study. One region, Amhara, is a highland area dominated by mixed crop-livestock systems while the remaining regions are all pastoral areas. The prices of vaccine (0.49 Ethiopian birr \[ETB\]/dose) and reconstituting saline (0.04 ETB/dose) were the same for all regions as they were purchased from the same source, the National Veterinary Institute in Bishoftu. The costs associated with the different components of delivery for each examined region is shown in [Fig. 2](#fig0010){ref-type="fig"} with tables presented in Appendix C. Vaccine transport and storage costs were similar among the different regions but differences were seen for the other parameters most noticeably for the farmer time component in Amhara. Field delivery costs were highest in the SNNP where there was minimal expenditure on publicity and mobilisation.Fig. 2Different costs associated with different cost components in four regions of Ethiopia or PPR vaccination (all costs in Ethiopian birr, ETB).Fig. 2
Observations of the vaccination in the field indicated that a proportion of the attempted vaccinations failed as the needle was not applied to the animal and the vaccine was simply ejected onto the animal's coat or floor. [Fig. 3](#fig0015){ref-type="fig"} shows the percentage of vaccination shots that were missed due to inadequate application at the herd/flock level ranged from 0 to 33% (overall mean across all four locations was 12.2%). [Fig. 4](#fig0020){ref-type="fig"} incorporates the cost of vaccine and associated wastage through missed shots and discards. The cost for the overall vaccine cost (vaccine, delivery and opportunity costs of farmers time) was approximately 6 ETB per animal in Amhara compared to approximately 3 ETB in the other areas indicating that vaccination costs in mixed crop livestock systems are approximately double that in pastoral areas of Ethiopia. A summary of the costs per animal in the four different locations is shown in [Fig. 4](#fig0020){ref-type="fig"} with a range of just over 3 ETB per animal in SNNP and just over 6 ETB in the mixed crop livestock system in Amhara.Fig. 3Frequency distribution of the mean proportion of PPR vaccinations missed from data collected at four different sites in Ethiopia.Fig. 3Fig. 4Total mean costs for each animal vaccinated for PPR in four regions of Ethiopia incorporating the cost of vaccine and delivery accounting for wastage through missed doses and discarded partially used bottles in each area (all costs in Ethiopian birr, ETB).Fig. 4
Differences were observed in the contribution of the different cost components to the overall costs in the different systems. In pastoral areas, the main cost was the field delivery whilst farmer's time was the largest contributor in the mixed-crop livestock system ([Table 1](#tbl0005){ref-type="table"}, [Fig. 5](#fig0025){ref-type="fig"}).Table 1Relative percentage of different cost components contributing to the mean vaccination cost for PPR among Pastoral and Mixed-Crop livestock systems in Ethiopia. Costs in Ethiopian birr (ETB).Table 1ComponentMean cost (ETB) per dose (%)PastoralMixed Crop LivestockVaccine0.53 (20.2)0.53 (10.3)Vaccine transport0.031 (1.0)0.022 (0.36)Vaccine storage0.00025 (0.0081)0.00022 (0.0036)Field delivery2.0 (64.8)1.1 (17.3)Farmer's time0.22 (7.2)4.0 (65.0)Co-ordination0.17 (5.6)0.29 (4.7)Publicity0.041 (1.3)0.14 (2.4)Fig. 5Percentage contribution of various components to PPR vaccination costs in pastoral and mixed crop livestock systems in Ethiopia.Fig. 5
[Fig. 6](#fig0030){ref-type="fig"} presents the national level costs based on the parameters given in [Table 2](#tbl0010){ref-type="table"}. The median estimate from the simulation estimated the cost in the pastoral system as US\$1.0 million (range 0.57--1.9) compared to \$US3.0 million (range 1.1--15.3) in the mixed crop livestock system assuming 80% vaccination coverage in both systems in a one year period.Fig. 6Population level costs (in \$US) for PPR vaccination for a single year in Ethiopia at 80% coverage of all small ruminants.Fig. 6Table 2Parameters used for population level stochastic model to estimate national PPR vaccination costs in Ethiopia based on collected field data.Table 2ParameterPastoralMixed-crop livestockDistributionValuesDistributionValuesVaccine transportUniformMin: 0.030\
Max:0.032--0.022Vaccine storageUniformMin: 0.00042\
Max: 0.00050--0.0020Field deliveryGammaRate: 7.3\
Shape: 3.6--1.1Farmers timeExponentialRate: 5.5ExponentialRate:0.25Co-ordinationUniformMin: 0.11\
Max:0.22--0.29PublicityUniformMin: 0\
Max: 0.11--0.14Missed shotsExponentialRate: 10.3ExponentialRate: 10.3Vaccine discardExponentialRate: 6.8ExponentialRate: 6.8Proportion adult animals--0.6--0.6Proportion young animals--0.4--0.4Adult vaccine coverage--0.8--0.8Young vaccine coverage--0.8--0.8
4. Discussion {#sec0035}
=============
The results presented are the first published account of PPR vaccine delivery costs based on field data incorporating vaccine wastage and opportunity costs for farmers. The average vaccination cost per dose was approximately 6 and 3 ETB in mixed crop-livestock and pastoral systems respectively. Based on the exchange rate at the time of writing, this equates to around 0.2 and 0.1 US\$ respectively. An unpublished rapid evaluation of costs by consultants estimated values of 1.13 (0.04US\$) and 2.12 ETB (0.078 US\$) in the pastoral Somali region when performed by private and public veterinary services respectively. Unlike the results in this study, these latter estimates did not consider the impact of opportunity cost of farmer's time, vaccine storage, or the cost of the vaccine (incorporating vaccine wastage through missed shots or residual wastage in vials), which likely accounts for the lower observed costs.
The estimates from Tago et al ([@bib0045]) based on a model parameterised with data from Senegal ranged from 0.20 USD (95%CI 0.18-0.21) to 0.34USD (95%CI 0.24-0.54) per dose depending on the assumptions made on number of animals vaccinated per day. These estimates are similar to those reported in this study although they did not consider the opportunity costs of the farmers and the model was mostly parameterised from expert opinion. In their model, they highlighted the importance of team "productivity" in terms of numbers of animals vaccinated per day. The assumed number of doses administered per team was lower than that recorded in the current study, even in the high productivity scenario. Related to vaccinator productivity, it is worth highlighting that vaccination is not the same as immunization, and a better evaluation would be based on serological outcomes rather than simply numbers of animals vaccinated.
The greatest proportion of the cost varied in the different production systems. In the pastoral system it was field delivery, while in the mixed crop-livestock system it was farmer's time. Tago et al ([@bib0045]) also reported the relative contribution of different cost components. Under high productivity scenarios the largest contributor was "injection supplies" followed closely by the cost of vaccine and staff. In the low productivity scenarios, staff time was over half of the cost. Farmer time was not considered in their analysis but has important implications as it represents an opportunity cost in terms of revenue forgone by the farmer in participating in the vaccination campaign. This component differed greatly in the production systems at approximately 65% in the mixed crop livestock system and 6% in the pastoral system. The actual time invested by farmers was equivalent in the two systems but the smaller sizes of flocks/herds in the mixed crop livestock system means the farmer has to spend relatively more time per animal owned. Flock/herd sizes tend to be lower in mixed crop livestock systems so this aspect could potentially compromise participating in vaccination campaigns aimed at high coverage to reduce the impact and circulation of PPR virus. The cost of farmer's time per hour was based on the lowest possible value of labour. This cost is expected to vary with season and may represent an over or underestimate depending on the seasonal demands in farming. Collecting data to inform this parameter would be useful in future studies on livestock vaccination as it helps to understand which people will be most inconvenienced in the short term by vaccination campaigns.
A major objective of this work was to develop a framework for collecting data from vaccination campaigns that could be applied elsewhere and adapted for other livestock diseases. Data collection was not standardised at the beginning of the project leading to limitations in the work. These include inconsistent measurements of vaccine wastage through missed shots and insufficient records of disposal of residual vaccine when moving between sites. These limitations highlight areas that should be accounted for with greater certainty in future economic studies. However, the detailed observations made in Afar and SNNP raises significant concerns over vaccine wastage that could have an important impact on national and global cost estimates for PPR vaccination. The unexpectedly high rate of missed shots is likely related to a lack of handling facilities that were not present in any campaign area. Vaccination was delivered either into the neck or medial aspect of the thigh by lifting of the hind limbs. These wastages could be reduced with: different application methods; improved training of staff; and access to adequate equipment for vaccination and animal handling. The improved effectiveness for these different interventions needs to be tested and costed in order to see what is the more appropriate. Moreover the vial size of 100 ml at 1 ml per dose could be leading to significant wastage in systems with small herd sizes ([@bib0005]). Experience during the rinderpest campaign was that a lower number of doses per vial did not introduce a cost saving due to the relatively higher packaging costs per dose (Gijs Van 't Klooster, Personal Communication) although an economic analysis based on small ruminant systems would be worthwhile.
The present study did not attempt to assess other negative impacts of vaccination. Anecdotally, farmers report adverse events such as abortion although this has not been quantified. Such adverse events could have an impact on coverage which again is more likely in smaller flocks/herds where farmers may be less inclined to take a risk with small numbers of animals, and there is also likely a seasonal effect to the perceived risk. Vaccination campaigns should be followed with more rigorous evaluation of adverse events to understand their impact, how they may be mitigated, and the effect they may have on coverage in subsequent vaccination campaigns.
PPR is a highly transmissible infection and consequently high vaccine coverage is required to control disease at a population level. This study did not attempt to quantify the cost of an entire campaign that will include factors such as background epidemiological studies to identify high-risk areas for vaccination, post-vaccination monitoring and surveillance to evaluate the impact of the campaign. Although the vaccine is thought to provide life-long protection, the high turnover of the population creates a unique challenge for maintaining vaccination coverage that is likely to be much greater than for rinderpest with different costs involved. This study developed a framework for estimating costs of vaccination campaigns for a disease identified by OIE and FAO for global eradication. Data collection and analysis identified important factors that could be having a significant impact on vaccination costs and willingness of farmers to participate. Further studies are required based on the established framework which should be incorporated into all publically funded vaccination campaigns to ensure efficient allocation of limited resources.
Appendix A. Supplementary data {#sec0045}
==============================
The following are Supplementary data to this article:
This project was funded in part by the Bill & Melinda Gates Foundation and with UK aid from the UK Government through GALVmed. Dr Lyons also acknowledges funding from a Biotechnology and Biological Sciences Research Council (BBSRC) funded fellowship (grant code: BB/E/I/00007004). The findings and conclusions contained within are those of the authors and do not necessarily reflect positions or policies of the Bill & Melinda Gates Foundation, the UK Government or BBSRC. The authors would also like to thank Dr Gijs Van 't Klooster for his input into the project.
Supplementary material related to this article can be found, in the online version, at doi:<https://doi.org/10.1016/j.prevetmed.2018.12.007>.
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the paper and its Supporting Information files.
Introduction {#sec001}
============
Mobile games are very diverse, and each game has its own characteristics. The games that players become extremely passionate about have some key attractive features. These games induce different emotions and sensory stimulations in order to keep players interested in playing the game. This study collected related papers evaluating games in the literatures, described below.
Lu and Wang \[[@pone.0216276.ref001]\] explored the role that addiction to online games plays in the relationship between online satisfaction and loyalty.
Mohamed and Jaafar \[[@pone.0216276.ref002]\] used a comprehensive evaluation method to evaluate elements of fun and education in games. Chu\[[@pone.0216276.ref003]\] evaluated player experience of games through a playability matrix, which was constructed from an analysis of the literature. In the study of Chen, Shen \[[@pone.0216276.ref004]\], "ease and convenience", the design features of social network games in usability were found through the methods of Kansei Engineering and the concrete features included "having fun instantly" and "playing anywhere via the internet". Grodal \[[@pone.0216276.ref005]\] researched how games evoked players' much stronger emotions. Yee \[[@pone.0216276.ref006]\] also studied the emotional investment of players and argued that the large amount of time spent by MMORPG players implies that they invest a significant level of emotion in playing this game. Lin, Hung \[[@pone.0216276.ref007]\] utilized focus group interview and statistical analysis of network survey to find that driving highly involved MMORPG gamers is coming from six main reasons: satisfaction, adventure, victory, socialization ability, self-actualization, and advancement of wealth and status. Cota et al.\[[@pone.0216276.ref008]\] applied a questionnaire and interviews to find that motivated the elderly to play mobile games was to help in the treatment of cognitive disorders due to aging.
Most papers in the literature have focused on researching players' emotions or perceptions \[[@pone.0216276.ref001], [@pone.0216276.ref002], [@pone.0216276.ref009], [@pone.0216276.ref010]\] and have scarcely explored player behaviors by focusing on games' concrete design features.
Hence, this study specially selects music games on mobile devices as a case for evaluating games, trying to extract design features of music games through a series of research processes. In general, games in this genre typically challenge players' sense of rhythm. To achieve high scores, a player has to press buttons at precise times in a sequence shown on the screen. Compared with other types of games, music games contain no violence. They are entertaining, educational\[[@pone.0216276.ref011]\], and easy to operate, and therefore attract a wide range of players, so they are worth studying in detail.
This study proposes a hypothesis that different design features can attract player population with different level of involvement, and designers must precisely grasp these features when developing mobile games. The objective of this paper is to explore the following two parts: 1. The design features of music games on mobile devices; 2. The cognition and preference of the different player groups toward these design features.
Methods {#sec002}
=======
In order to clarify the relationship between players' abstract emotions and concrete design features, and to illustrate the cognition difference between different involvement players toward games, this study mainly utilized three methods to collect and analyze data.
[Fig 1](#pone.0216276.g001){ref-type="fig"} shows the research process, which indicates EGM and CIP measures were performed prior to Multiple Regression Analysis with different samplings in order to achieve two research objectives.
{#pone.0216276.g001}
This study has been approved by Research Ethics Committee of National Taiwan University. The approval number is 201706ES011. The written forms of consent for interviews and questionnaire are obtained. The details of all methods are explained as follows.
1. Evaluation Grid Method {#sec003}
-------------------------
This study utilized EGM to extract concrete design features of music games. EGM is to collect and organize data with the methods of in-depth interviews and drawing of diagrams. It is one of research methods to transform players' emotions into concrete design features of games.
### (1) About Evaluation Grid Method (EGM) {#sec004}
Evaluation Grid Method, derived from the field of psychology, is an important research method in Miryoku Engineering to capture personal cognitive concepts and organize them. This method explicitly explores similarities or differences of objects through personal interviews with comparisons of objects, and then analyzes the characteristics of the targets \[[@pone.0216276.ref012]--[@pone.0216276.ref014]\]. Sanui\[[@pone.0216276.ref015]\] further developed this method. First, in the assessment of a target, the interviewee needs to answer the reasons of his likeness to this object.
In the related studies of Miryoku Engineering, EGM provides an approach based on theories to analyze design features of products. In order to get consumers' feelings to the attractiveness of products, in-depth interviews are conducted with stimuli provided in the topics and comparisons of likeness which make obvious difference in interviewees' feelings. In this way, the researcher will extract the original concepts of the topics from the interviewees and then guide the interviewees to analyze their concepts more clearly so that the concepts can be transformed into abstract emotions and concrete features for organizing real thoughts of the interviewees \[[@pone.0216276.ref016]\].
In other words, the process is conducted hierarchically. The manipulation steps are explained as follows:
1. First, an interviewee is given stimuli, such as oral descriptions and comparisons through pictures. The interviewee is asked to evaluate what is good or bad and what he/she likes or dislikes about the products.
2. The interviewee is asked to describe the reasons for the preference in his/her own words, and these reasons are recorded as "original evaluation items" or "middle evaluation items."
3. The evaluation responses are recorded and the interviewee is asked for the more abstract meanings and emotions behind a response such as "upper evaluation items."
4. We refer to the evaluation items and continue asking the interviewee for more concrete formative conditions and features of the reasons for labeling items as "lower evaluation items".
5. All evaluation items are compiled into a diagram. The upper items are on the left, the middle ones in the middle and the lower ones on the right. All items are connected with straight lines to indicate the hierarchical relationships.
In the literature, researchers have utilised the EGM to investigate individual's emotions and cognitions for objects and extract the exact inner decision-making factors influencing each individual in several fields, such as the design\[[@pone.0216276.ref016]--[@pone.0216276.ref022]\], interactive media\[[@pone.0216276.ref004], [@pone.0216276.ref023]--[@pone.0216276.ref026]\], advertisement and marketing\[[@pone.0216276.ref016], [@pone.0216276.ref027]--[@pone.0216276.ref029]\], education and psychology\[[@pone.0216276.ref030]\], and tourism\[[@pone.0216276.ref031], [@pone.0216276.ref032]\] fields.
In addition to exploring the emotions mobile games can invoke in players, which has been performed in previous researches, this study aims to transform these emotions into concrete design features with a systematical method, or, in other words, to discover successful features of games. The results can be used as a reference for game designers in the future.
### (2) The operating process {#sec005}
This study recruited 10 researchers who had playing experience in music games and had received research interview training; 6 were male, and the researchers were aged between 23 and 41 years. They were employed to find frequent users of music games in Taiwan to interview. The requirement to be frequent users for interviews is playing any music game every week for more than half year.
First, the interviewees were asked to recall and talk about their experiences playing music games. They then opened the game app using a mobile device for reviewing the game design features. The actual operation can avoid flaws of biasing caused from that interviewees merely relied memory to answer questions\[[@pone.0216276.ref033]\]. Then, the interviewees were asked to talk about the elements attracted them.
For example, the interviewers asked, 'Compared with other games, which design features in this game attract you?' After the interviewees had responded, the interviewers continued to ask about specific reasons and the hidden, abstract items regarding the interviewees' feelings.
An experienced interviewer would induce the interviewee to express more deep-seated preferences and feelings under the objective conditions of not deliberately guiding the interviewee\[[@pone.0216276.ref034]\].
Because data were collected using in-depth interviews, which are time consuming, a large-scale survey was not feasible\[[@pone.0216276.ref017], [@pone.0216276.ref029]\]. This study interviewed 22 frequent users of music games; 11 were male, 11 were female, and they were aged between 20 and 45 years.
Next, in-depth interview results related to the music games were compiled into a diagram. The upper items are on the left, the middle items in the middle, and the lower items on the right. Items connected with straight lines indicate hierarchical relationships. This arrangement constituted the EGM diagram, which integrates the data, and determines the relationships among the upper items (abstract emotions), middle evaluation items (original evaluation items), and lower evaluation items (concrete design features), which is presented in [Fig 2](#pone.0216276.g002){ref-type="fig"}.
{#pone.0216276.g002}
2. CIP measures {#sec006}
---------------
This study adopted the consumer involvement profile proposed by Kapferer and Laurent\[[@pone.0216276.ref035]\] as a basic form to measure involvement and designed questions coordinating the characteristics of mobile games to determine groups of different involvement. Finally, we considered that the concrete design features may have different significance and meaning to players with different levels of involvement in different game types.
Consumer involvement profiles measure the level of relevance between games and players through questions, and a 5- point Likert scale was adopted to assess the following 13 questions regarding level of involvement:
1. Playing mobile games is important in my life.
2. Playing mobile games makes me feel happy and contented.
3. I can be very anxious if I cannot find a phone or tablet to play a game.
4. I am very anxious if I cannot find a mobile game I like.
5. I often pay to purchase items or points in mobile games.
6. I spend more than an hour per day playing mobile games.
7. I pay attention to news related to mobile games.
8. I know all the functions and characteristics of the mobile games I play.
9. Playing mobile games gives me a sense of achievement.
10. Getting a high score, rankings, and items in a game are important to me.
11. I love discussing the topic of playing mobile games with my friends.
12. I care what people think about the fact that I play mobile games.
13. I can find out other people's preferences from the mobile games they play.
This study grouped involvement levels using the quartile method based on the players' total scores in assessment questions of involvement. The players were divided into three groups: high involvement (first 25%), medium involvement (middle 50%), and low involvement (last 25%). The total score was obtained from the degree of the subjects' agreement in every assessment question (1 to 5). Every subject's 13 scores were summed.
3. Multiple regression analysis {#sec007}
-------------------------------
In the first method, aforementioned EGM obtained original evaluation items and concrete design features. We converted these into a questionnaire with measurable questions and that could be distributed to a large number of game players. We used a 5-point Likert scale to assess the subjects' degree of agreement.
For examples: "The games have favourable level design because of providing a teaching mode"; "The games have favourable level design because of ability to choose difficulty or speed of levels". The subjects gave 1\~5 points for all original evaluation items and concrete design features according to their agreement.
Besides, in the same questionnaire, this study used the second method, CIP measures, to identify three involvement groups which each subject was belong to.
Finally, this study used the third method, statistics of multiple regression analysis\[[@pone.0216276.ref036]\] via SPSS, to explore how the variables of the three involvement groups are affected by the variables of concrete design features of music games.
Results {#sec008}
=======
1. Results of EGM {#sec009}
-----------------
This study individually produced EGM diagrams for the 22 interviewees' data, and then integrated the 22 diagrams into one, showed in [Fig 3](#pone.0216276.g003){ref-type="fig"}. The integrated EGM diagram showed 44 concrete design features and 6 original evaluations items. We can clearly see why the interviewee is attracted by a particular event after completing this step.
{#pone.0216276.g003}
These items were obtained using qualitative research methods. However, this study also required quantitative research methods to verify these items. Therefore, we converted 44 concrete design features and 6 original evaluations items into a questionnaire with measurable questions; 210 valid questionnaires were collected.
2. Results of questionnaire {#sec010}
---------------------------
### (1) Demographic characteristics {#sec011}
The construct validity of the 44 measurable questions was analysed. The Kaiser--Meyer--Olkin (KMO) measure obtained was 0.92, and the Bartlett's test of sphericity score was 5094.96, which indicated relatively high validity. In the reliability analysis, the Cronbach's alpha was 0.95, indicating that the questionnaire had high reliability. The demographics of the respondents are presented in [Table 1](#pone.0216276.t001){ref-type="table"}.
10.1371/journal.pone.0216276.t001
###### Socio-demographic information.
{#pone.0216276.t001g}
Number
----------------------- ------------------- ---------
Gender Male 127
Female 83
Age 10--19 147
20--29 50
30--39 8
40--49 5
Education Elementary school 33
Junior high school 2
Senior high school 82
College or University 89
Graduate school 4
Profession Student 190
Government employee 15
Service 4
Manufacturing 1
**Total** **210**
### (2) Grouping according to involvement levels {#sec012}
The total score was obtained from the degree of the subjects' agreement in every assessment question (1 to 5). Every subject's 13 scores were summed. The highest sum obtained for the questionnaire was 61 points, and the lowest sum was 13. The involvement grouping results are shown in [Table 2](#pone.0216276.t002){ref-type="table"}.
10.1371/journal.pone.0216276.t002
###### Grouping of involvement.
{#pone.0216276.t002g}
Sum of score Involvement Number of subjects
-------------- --------------------------- --------------------
**13--25** *LI* (Low involvement) 51
**26--42** *MI* (Medium involvement) 105
**43--61** *HI* (High involvement) 54
### (3) Results of multiple regression analysis {#sec013}
In EGM, an original evaluation item is a goal those concrete design features have to achieve. For example of the first evaluation item in [Fig 4](#pone.0216276.g004){ref-type="fig"}, in order to make players feel "favourable level design", designers can strengthen the features, "the teaching mode", "the ability to choose difficulty or speed of levels", "the ability to replay music or restart levels", and so on.
{#pone.0216276.g004}
In the perceptions of these three involvement groups, is there any design feature which has linear relationship in these 6 original evaluation items? In another word, to improve these six items efficiently, which design feature do designers have to strengthen?
This study conducted multiple regression analysis separately with the three groups to determine how the three involvement groups were affected by 44 concrete design features in the six original evaluation items. In the multiple regression analysis, the dependent variables were subjects' scores (1--7) of the importance of the original evaluations items in the questionnaires, and the independent variables were their degree of agreement (1--5) for each concrete design feature (see [Fig 4](#pone.0216276.g004){ref-type="fig"}). There were six items and three involvement groups; thus, 18 multiple regression analyses were performed in total (6 items \* 3 groups). The multiple regression analysis results for the three groups are shown in Tables [3](#pone.0216276.t003){ref-type="table"}, [4](#pone.0216276.t004){ref-type="table"} and [5](#pone.0216276.t005){ref-type="table"}, respectively.
10.1371/journal.pone.0216276.t003
###### Multiple regression analysis with *LI*.
{#pone.0216276.t003g}
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Original Evaluations Items Concrete Design Features with Significant Impact Constant B Value Beta\ T Value Sig. Mean/SD
(β)
--------------------------------------------- ------------------------------------------------------------ ---------- --------- --------------------------------------------- --------- ----------------------------------------------- ---------
***Item 1*. *Favourable Level Design*** E. Increasing difficulty or speed of the levels 1.945 .788 .482 3.846 .000[\*\*\*](#t003fn003){ref-type="table-fn"} 3.014\
1.236
***Item 2*. *Favourable Artistic Design*** C. Soft and coordinated colours .974 1.026 .789 8.982 .000[\*\*\*](#t003fn003){ref-type="table-fn"} 3.087\
1.314
***Item 3*. *Favourable Interface Design*** A. Ease of control by clicking or sliding a finger .789 .557 .415 3.240 .002[\*\*](#t003fn002){ref-type="table-fn"} 3.168\
1.253
G. Large buttons .492 .375 2.292 .005[\*\*](#t003fn002){ref-type="table-fn"}
***Item 4*. *Favourable Reward Design*** B. Extra games to earn more points after completing levels 1.227 .881 .138 6.387 .000[\*\*\*](#t003fn003){ref-type="table-fn"} 3.181\
1.278
***Item 5*. *Favourable Music Design*** C. Original songs created by well-known composers 1.878 .841 .643 5.881 .000[\*\*\*](#t003fn003){ref-type="table-fn"} 3.302\
1.278
***Item 6*. *Playable at odd moments*** C. Designs playable without Internet 2.345 .772 .611 5.402 .000[\*\*\*](#t003fn003){ref-type="table-fn"} 3.480\
1.324
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\*P \< 0.0
\*\* P \<0.0
\*\*\* P \< 0.001
10.1371/journal.pone.0216276.t004
###### Multiple regression analysis with *MI*.
{#pone.0216276.t004g}
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Original Evaluations Items Concrete Design Features with Significant Impact Constant B Value Beta\ T Value Sig. Mean/SD
(β)
---------------------------------------------------------------- ------------------------------------------------------------------ ---------- --------- --------------------------------------------- --------- ----------------------------------------------- ---------
***Item 1*. *Favourable Level Design*** E. Increasing difficulty or speed of the levels 3.108 .432 .328 3.511 .001[\*\*](#t004fn002){ref-type="table-fn"} 3.668\
1.028
J. Ability to download other music to play game .189 .187 1.998 0.048[\*](#t004fn001){ref-type="table-fn"}
***Item 2*. *Favourable Artistic Design*** E. High recognition in artistic styles 2.513 .418 .304 3.314 .001[\*\*](#t004fn002){ref-type="table-fn"} 3.656\
1.042
F. Cute characters .323 .292 3.184 .002[\*\*](#t004fn002){ref-type="table-fn"}
***Item 3*. *Favourable Interface Design*** F. One-dimensional progress of rhythm graphics 2.899 .321 .254 2.491 .014[\*](#t004fn001){ref-type="table-fn"} 3.655\
0.949
B. Designs that prompt rhythm points using glowing or flashing .316 .239 2.343 .021[\*](#t004fn001){ref-type="table-fn"}
***Item 4*. *Favourable Reward Design*** E. Characters grow or become stronger after a level is completed 1.963 .533 .399 4.574 .000[\*\*\*](#t004fn003){ref-type="table-fn"} 3.724\
0.990
B. Extra games to earn more points after completing levels .330 .233 2.668 .009[\*\*](#t004fn002){ref-type="table-fn"}
***Item 5*. *Favourable Music Design*** A. Diverse music styles .746 .691 .481 5.637 .000[\*\*\*](#t004fn003){ref-type="table-fn"} 3.823\
1.049
G. Music with powerful rhythms .309 .233 2.614 .010[\*](#t004fn001){ref-type="table-fn"}
E. Funny and cute music .229 .186 2.253 .013[\*](#t004fn001){ref-type="table-fn"}
***Item 6*. *Playable at odd moments*** B. Ability to continue paused game anytime 1.654 .507 .320 3.356 .001[\*\*\*](#t004fn003){ref-type="table-fn"} 3.998\
1.016
C. Designs playable without Internet .421 .304 3.195 .002[\*\*](#t004fn002){ref-type="table-fn"}
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\* P \< 0.0
\*\* P \<0.01
\*\*\* P \< 0.001
10.1371/journal.pone.0216276.t005
###### Multiple regression analysis with *HI*.
{#pone.0216276.t005g}
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Original Evaluations Items Concrete Design Features with Significant Impact Constant B Value Beta\ T Value Sig. Mean/SD
(β)
--------------------------------------------- ------------------------------------------------------------ ---------- --------- ------------------------------------------- --------- ----------------------------------------------- ---------
***Item 1*. *Favourable Level Design*** G. Rhythm notes with diverse actions 2.481 .455 .273 2.105 .04[\*](#t005fn001){ref-type="table-fn"} 4.029\
0.993
A. Teaching mode .316 .261 2.015 .049[\*](#t005fn001){ref-type="table-fn"}
***Item 2*. *Favourable Artistic Design*** B. Gorgeous special effects 2.656 .674 .413 3.273 .002[\*](#t005fn001){ref-type="table-fn"} 3.962\
1.011
***Item 3*. *Favourable Interface Design*** A. Ease of control by clicking or sliding a finger 1.809 .886 .528 4.484 .000[\*\*\*](#t005fn002){ref-type="table-fn"} 3.881\
0.967
***Item 4*. *Favourable Reward Design*** B. Extra games to earn more points after completing levels 3.466 .521 .438 3.515 .001[\*\*\*](#t005fn002){ref-type="table-fn"} 3.984\
1.032
***Item 5*. *Favourable Music Design*** C. Original songs created by well-known composers 4.386 .432 .325 2.475 .017[\*](#t005fn001){ref-type="table-fn"} 4.082\
0.957
***Item 6*. *Playable at odd moments*** C. Designs playable without Internet 2.949 .609 .552 4.773 .000[\*\*\*](#t005fn003){ref-type="table-fn"} 4.171\
0.945
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\* P \< 0.05
\*\* P \<0.01
\*\*\* P \< 0.001
Discussion {#sec014}
==========
The schedule and manpower of this study are finite, so there are the following research limitations:
1. When we interviewed a frequent user, he/she could only provide ideas or feelings for his/her familiar games. This study has sought as many interviewees as possible, but still may not be able to cover all music games.
2. VanVoorhis and Morgan \[[@pone.0216276.ref037]\] recommended an absolute minimum of 10 participants per predictor variable for regression using six or more predictors. In this study, number of predictors is 44, so the ideal number of the subjects is supposed to be 440. However, the subjects of the questionnaires must be the players who had used music games, which needed to be screened strictly and the number was limited. Moreover, according to Cohen \[[@pone.0216276.ref038]\], we calculated the number of samples using the medium effect size f^2^ = 0.15, α = 0.05, and maintaining the power = 0.8, and determined the number of subjects needed for the questionnaire was about 210 ultimately.
This section presents explanation of the six original evaluation items in EGM and discusses the common concrete design features that different involvement groups have significance.
1. The results of EGM {#sec015}
---------------------
The EGM diagram in [Fig 3](#pone.0216276.g003){ref-type="fig"} displays the detailed results of the interviews.
### (1) Favourable level design {#sec016}
In this item, there were 11 concrete design features in which most players expressed a preference, including 'A. Teaching mode', 'B. Ability to choose difficulty or speed of levels', 'C. Ability to replay music or restart levels', etc. The abstract reasons in this item were 'surprise', 'variety', 'challenge', etc.
These results agree with the design principles of casual games suggested by Greechan\[[@pone.0216276.ref039]\]. Game play complexity is carefully designed and gradually introduces players to higher levels. Greechan also mentioned that forgiving, nonpunishing gameplay provides numerous hints before a punishment is given and encourages players to continue attempts to increase their sense of accomplishment. In addition, various ways of clicking with the rhythm and designs that enable players to choose levels of difficulty or speed make games flexible and open. Music games are thus similar to casual games in their level design, and these games should take players of different abilities into account using these methods.
### (2) Favourable artistic design {#sec017}
In this item, there were seven concrete design features in which most players expressed a preference, including 'A. Well-designed characters and scenes', 'B. Gorgeous special effects', 'C. Soft and coordinated colours', etc. The abstract reasons in this item were fine images, gorgeous special effects, and beautiful backgrounds, showing rich content to create visual excitement, portray high quality, and offer comfort.
### (3) Favourable Interface Design {#sec018}
In this item, there were 10 concrete design features for which most players expressed a preference, including 'A. Ease of control by clicking or sliding a finger', 'B. Designs that prompt rhythm points by glowing or flashing', 'C. Fixed graphics to denote rhythm', etc. The abstract reasons in this item were 'challenge', 'tension', 'easy operation', etc. These indicate that the interface features of attractive music games include the following: (a) glowing or flashing rhythm notes are clear and easy to identify, (b) illustration not text for game descriptions, and (c) advertisements in games are few. These features generate relaxing visual effects for game players and facilitate game playing.
### (4) Favourable reward design {#sec019}
In this item, there were eight concrete design features for which most players expressed a preference, including 'A. Introduction of music knowledge', 'B. Progress of story or missions after completing levels', 'C. Extra games to earn more points after completing levels', etc. The abstract reasons in this item were 'expectation', 'interactivity', 'rich playability', etc. These include internal and external rewards:
1. Internal rewards: Scores, gold coins, mileage, health points, and lucky draw mechanisms are the internal rewards in general games and are also effective rewards in music games. Making characters stronger and progressing the story or mission are also features of music games. In the upper items, these mechanisms prompt players to feel a sense of accomplishment at their overcoming of challenges, and storytelling factors are also crucial characteristics of some music games.
2. External rewards refer to players' competition over rankings, and rankings are even shared with friends in the community. This result agrees with those of Campbell\[[@pone.0216276.ref040]\] and Radoff\[[@pone.0216276.ref041]\] regarding game communities attracting players to discuss and share their game experiences, achievements, and photos with friends.
### (5) Favourable Music Design {#sec020}
In this item, there were seven concrete design features for which most players expressed a preference, including 'A. Diverse music styles', 'B. Popular songs such as those from movie soundtracks', 'C. Original songs created by well-known composers', etc. The abstract reasons in this item were 'interest', 'rich playability', 'accomplishment', etc. Players prefer to participate in games with diverse familiar music of different styles, which produces pleasure and a high sense of accomplishment. Listening pleasure obtained from loving and enjoying music is the main value of music games.
### (6) Playable at odd moments {#sec021}
In this item, there were three concrete design features for which most players expressed a preference, including 'A. Level can be completed in 5 minutes', 'B. Ability to continue paused game anytime', and 'C. Playability without internet'. The abstract reasons in this item were 'easy operation', 'killing time', 'relaxation', etc.
Music games are easy to start, pause, leave, and return to\[[@pone.0216276.ref042]\] because they generally do not require Internet connectivity. Thus, players can choose the time in which they play music games. Besides, the length of a song is generally only 3--5 minutes, so players can quickly know whether they pass levels or not. Therefore, it is critical to allow players short bursts of enjoyment.
2. The results of statistical analysis {#sec022}
--------------------------------------
Tables [3](#pone.0216276.t003){ref-type="table"}, [4](#pone.0216276.t004){ref-type="table"} and [5](#pone.0216276.t005){ref-type="table"} were combined to obtain [Table 6](#pone.0216276.t006){ref-type="table"}, which compares the multiple regression results of *HI*, *MI* and *LI*. [Table 6](#pone.0216276.t006){ref-type="table"} shows the concrete design features having significant levels of impact on different involvement groups.
10.1371/journal.pone.0216276.t006
###### Comparison of the concrete design features reaching significant levels to impact different involvement groups.
{#pone.0216276.t006g}
Original Evaluation Items *HI* *MI* *LI*
--------------------------------------------- ------------------------------------------------------------------ ------------------------------------------------------------ ------------------------------------------------------------
***Item 1*. *Favourable Level Design*** G. Rhythm notes with diverse actions
A. Teaching mode
E. Increasing difficulty or speed of the levels E. Increasing difficulty or speed of the levels
J. Ability to download other music to play game
***Item 2*. *Favourable Artistic Design*** B. Gorgeous special effects
F. Cute characters
E. High recognition in artistic styles
C. Soft and Coordinated colours
***Item 3*. *Favourable Interface Design*** A. Ease of control by clicking or sliding a finger A. Ease of control by clicking or sliding a finger
F. One-dimensional progress of rhythm graphics
B. Designs that prompt rhythm points using glowing or flashing
G. Large buttons
***Item 4*. *Favourable Reward Design*** B. Extra games to earn more points after completing levels B. Extra games to earn more points after completing levels B. Extra games to earn more points after completing levels
E. Characters grow or become stronger after a level is completed
***Item 5*. *Favourable Music Design*** C. Original songs created by well-known composers C. Original songs created by well-known composers
A. Diverse music styles
G. Music with powerful rhythms
E. Funny and cute music
***Item 6*. *Playable at odd moments*** C. Playable without Internet C. Playable without Internet C. Playable without Internet
B. Ability to continue paused game anytime
Two features reached the level of significance among all three groups: 'B. Extra games to earn more points after completing levels' in Item 4, and 'C. Playable without Internet' in Item 6.
The ability to play a game offline is crucial for players. Music games are easy to start, pause, leave, and return to\[[@pone.0216276.ref042]\] because they generally do not require Internet connectivity. Thus, players can choose the time in which they play music games, with them able to freely leave and return to the game. These two concrete design features, 'ability to continue paused games anytime' and 'playable without Internet' are concrete methods of enabling players relax when playing music games.
Some music games provide short extra games after the completion of levels, which allow players to earn more coins or points, giving players a sense of accomplishment and challenge. One feature significantly influenced the common cognition in the *MI* and *LI*: 'E. Increasing difficulty or speed of the levels' in Item 1.
Moreover, two features significantly influenced the common features in the *HI* and *LI*: 'A. Ease of control by clicking or sliding a finger' in Item 3, and 'C. Original songs created by well-known composers' in Item 5.
Conclusion {#sec023}
==========
This study discovered that only two concrete design features significantly influenced all three groups: 'Extra games to earn more points after completing levels' and 'Playable without Internet'.
Relevant literature on the involvement and online games has revealed that player groups at different levels of involvement have different perceptions towards the same product. The experiment conducted in this study yielded similar results. Groups with different levels of involvement in music games had different preferences of game design features.
This study not only investigated players' abstract emotions and perceptions, but also transformed those into concrete design features in game designs. These design features offer a concrete method for improving music games. Designers can strengthen specific design features according to their different targets in the gaming market.
Supporting information {#sec024}
======================
###### Original data for statistics.
(XLSX)
######
Click here for additional data file.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
| {
"pile_set_name": "PubMed Central"
} |
The content published in Cureus is the result of clinical experience and/or research by independent individuals or organizations. Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein. All content published within Cureus is intended only for educational, research and reference purposes. Additionally, articles published within Cureus should not be deemed a suitable substitute for the advice of a qualified health care professional. Do not disregard or avoid professional medical advice due to content published within Cureus.
Introduction
============
Iatrogenic dural puncture (IDP) usually presents as post-dural puncture headache (PDPH) caused by the sudden leak of cerebrospinal fluid (CSF). PDPH is defined as a headache that worsens on attaining upright position and is relieved quickly after resuming recumbent position \[[@REF1]\]. It can be frontal or occipital in location, radiating to neck/back/shoulder and is often accompanied by nausea, photophobia, and dizziness. Infrequently, IDP can have an atypical presentation \[[@REF2]\]. Here, we describe the case of a patient who developed intracranial hypotension, presenting as diplopia from abducens nerve palsy without the telltale signs of PDPH.
Case presentation
=================
A 37-year-old female with an oesophageal mass presented for distal oesopago-gastrectomy and roux-en-y oesophagojejunal anastomosis. For the planned operation, a T6-7 epidural was placed by the paramedian approach without complication for post-operative pain control. Ropivicaine 0.2% was started at 8 mL/hr with 4 mL boluses with a once-per-hour allowance. The patient underwent the planned procedure as expected.
On post-operative day one, the patient complained of severe incisional site pain. The epidural catheter was bolused without any relief of pain and sign of a sensory block. After discussion, the patient was offered a replacement of the epidural and she elected to proceed with the plan. There was noted difficulty in placing the epidural via the paramedian approach leading to an unintentional dural puncture on the first attempt. The epidural catheter was then successfully placed at T8-9 by a midline approach with improvement in incisional pain.
The day following the unintentional dural puncture, the patient complained of new headache, neck, and shoulder pain. The headache was not positional and no other symptoms of a PDPH were noted. The posterior headache along with neck and shoulder pain was described as debilitating. At this point, ultrasound-guided bilateral greater occipital nerve blocks were done with immediate relief of posterior head and neck pain, however, her shoulder pain persisted.
On post-operative day four, the epidural catheter was removed without any complications. Following the catheter removal, the patient reported cessation of shoulder pain. One week following the unintentional dural puncture, the patient complained of headache and double vision. She described the headache as dull, bilateral but worse on the left, and without a change in the intensity with positional changes. She denied hearing changes and photophobia. She complained of nausea that worsened with sitting or standing.
On physical exam, the patient had double vision on rightward gaze (divergence paresis), consistent with right cranial nerve 6 palsy. Intravenous hydration, fiorcet, and caffeine were ordered. Neurology service was consulted and based on their evaluation, concern was raised for microvascular etiology with a low suspicion for PDPH. Computed tomography angiography (CTA) of the head and neck was performed which was negative for carotid or vertebral dissection. In addition, magnetic resonance imaging (MRI) of the brain was performed and findings were consistent with intracranial hypotension (Figures [1](#FIG1){ref-type="fig"}-[2](#FIG2){ref-type="fig"}). Based on these findings, an epidural blood patch (EBP) was suggested. The EBP was placed at the lumbar spine and a total of 15 ml of blood was injected into the epidural space. Following the EBP, the patient's headache resolved and the cranial nerve palsy improved but persisted. Upon follow-up visit with the surgical services, there were no reported visual complaints.
{#FIG1}
{#FIG2}
Discussion
==========
Abducens nerve palsy presenting as diplopia is a rare but serious complication of intracranial hypotension from CSF leak. Among the cranial nerves causing ophthalmolplegia, the abducens or the sixth cranial nerve is the most frequently involved with horizontal diplopia and blurred vision \[[@REF3]\]. Zada et al. further mention that of the patients with cranial nerve palsy secondary to intracranial hypotension, 83% of patients had abducens (VI) nerve palsies, 14% had oculomotor (III) nerve palsies, and 7% had trochlear (IV) nerve paresis \[[@REF3]\]. The exact incidence of abducens nerve palsy is unknown but Thomke's institutional experience put the risk in a range of approximately 1 in 6,000 patients \[[@REF4]\]. Abducens nerve palsy due to intracranial hypotension is a benign condition and about 80% of patients recover spontaneously \[[@REF3]\].
The abducens nerve is most sensitive to intracranial hypotension. This is not due to its longer course as was earlier believed but due to three acute angulation points between the dural entrance point and its anastomosis with the periarterial sympathetic plexus \[[@REF5]\]. The diagnosis of IDP abducens nerve palsy is that of exclusion \[[@REF2]\]. Other etiologies including multiple sclerosis, myasthenia gravis, thyroid disease, diabetes, myositis, intracranial tumour, arteriovenous malformation, pachymeningitis, and intracranial hematoma should be ruled out \[[@REF6]-[@REF7]\].
The characteristic MRI findings are typical of intracranial hypotension consisting of small ventricles, diffuse post-contrast meningeal enhancement with downward displacement of the brainstem and subdural effusions. There could also be tearing of small meningeal vessels causing subdural hematoma \[[@REF8]\]. Diffuse pachymeningeal enhancement is one of the most common neuroimaging abnormalities in intracranial hypotension which can lead to suspicion of inflammation or neoplasm as the etiology. Clinical correlation is therefore extremely important \[[@REF6]\].
Literature search reveals both conservative and invasive therapies \[[@REF2],[@REF8]\]. Conservative treatment includes intravenous dexamethasone therapy followed by taper and lateral gaze exercises and/or alternate patching of the eyes every two hours for a couple of days followed by Fresnel prism eye lenses. Invasive treatment includes EBP, and in unresolved cases, strabismus eye repair \[[@REF5]\].
The timing of an EBP is unclear to derive from the existing literature \[[@REF8]-[@REF9]\]. However, there is an indication that the placement of EBP within 24 hours of presentation of abducens nerve palsy is most effective in treating the palsy by restoring the intracranial pressure back to normal \[[@REF2]\]. In a total of 27 patients, in two case reports where patients developed abducens nerve palsy after IDP, 25 patients had complete resolution of abducens nerve palsy from the EBP \[[@REF3],[@REF8]\].
The mechanism that accounts for the success of EBP is not completely understood. Presumably, when the blood is injected, it enters both the posterior and anterior epidural space \[[@REF10]\]. The thecal space is compressed and displaced by the blood. The blood also exits out of the intervertebral foramina and into the paravertebral space. 14 ml of blood can cause a mean spread up to six spinal segments cephalad and three segments caudad. Compression of the thecal space leading to elevation of subarachnoid pressure, may explain the rapid resolution of the headache \[[@REF10]\].
Conclusions
===========
Cranial nerve palsies from intracranial hypotension can occur even in the absence of the classic PDPH. The most common cranial nerve to be affected by intracranial hypotension is cranial nerve VI or abducens nerve. Unilateral abducens nerve palsy is more common than bilateral presentation. In the case of abducens cranial nerve palsy, headaches may precede the palsy, follow it, or maybe completely absent. This knowledge prevents unnecessary and dangerous workups including a diagnostic lumbar puncture. Cranial nerve palsies can present days after IDP clouding the diagnosis. Gadolinium-enhanced MRI is the investigation of choice in clinching the diagnosis of cranial nerve palsies following IDP. While there are no clear recommendations for when to place an EBP for treating the neurological symptoms, earlier treatment with EBP has a better prognosis.
The authors have declared that no competing interests exist.
Consent was obtained by all participants in this study
| {
"pile_set_name": "PubMed Central"
} |
\"Our poor family seems persecuted by this awful disease, the worst I know.\" Queen Victoria (1819-1901) of the United Kingdom of Great Britain and Ireland wrote in her diary. She is well known as the most famous carrier of hemophilia, the \"royal disease\", and had passed on the disease to several royal families in Europe through her daughters.
Since first described in a hemophilia patient by Lawrence and Johnson in 1941 \[[@B1]\], inhibitor development has been the most serious complication of hemophilia A treatment. Inhibitor development occurs in up to 36% of patients with severe hemophilia A \[[@B2], [@B3]\]. The inhibitor is an alloantibody against factor VIII (FVIII), which if developed during hemophilia treatment, impairs FVIII activity and thus neutralizes the effect of the \"factor replacement\" therapy, leading to increased complications and therapeutic cost \[[@B4]\]. Some genetic and environmental risk factors for inhibitor development are well known; however, it is still unclear why some, but not other patients, develop inhibitors. The inhibitors in hemophiliacs may be temporary or can be eradicated with immune tolerance induction (ITI) therapy. The recent study (International Immune Tolerance Study) of 115 \"good-risk\" subjects with severe hemophilia A and high-titer inhibitors showed that there was no difference between the low-dose (50 IU/kg 3 times/week) and high-dose (200 IU/kg/day) regimens in achieving tolerance, with the former taking longer \[[@B5]\]. These data might be important for nations with limited resources. Although the eradication of inhibitors and recommencement of FVIII replacement therapy is a long-term goal in treating inhibitor patients, ITI therapy has limited indications and a success rate of 63-80% \[[@B6]\].
Patients with high-titer inhibitors and intractable bleeding episodes should be given recombinant activated factor VII (r-FVIIa) or activated prothrombin complex concentrate (aPCC), called the hemostatic bypassing agent (to circumvent FVIII in the coagulation pathway), as the first choice therapy. The FEIBA NovoSeven Comparative (FENOC) Study evaluated the hemostatic efficacy of both products on 96 joint bleeds of 48 inhibitor patients and showed nearly equivalent efficacy, with both agents being effective and safe, although a substantial number of patients reported different efficacies for both agents \[[@B7]\]. Nevertheless, 10-20% of bleeding events in hemophiliacs with high-titer inhibitors cannot be controlled by either r-FVIIa or aPCC alone and thus may need sequential combined bypassing therapy (SCBT).
The rationale for the use of a combination of r-FVIIa and aPCC is based on data from experiments using a rabbit stasis model, where an early thrombotic effect was shown, indicating the synergistic effect of both agents on thrombus formation during stasis \[[@B8]\]. Clinically, SCBT is predicated upon the fact that neither bypassing agent controls bleeding completely and the response of each individual patient to either agent differs \[[@B9]\]. These aspects were also observed in the FENOC study \[[@B7]\], where up to 43.8% of subjects considered one product more effective than the other.
A recent study employing a comprehensive literature search revealed that a significant number of patients treated with SCBT experienced complications such as thrombosis: 5 of 9 patients with acquired hemophilia and 5 of 40 with congenital hemophilia, among which 4 cases were fatal \[[@B8]\]. Moreover, no randomized controlled study has been carried out to investigate SCBT. Hence, the sequential use of both bypassing agents should have strict indications as mentioned earlier and should be accompanied by a thorough follow up to assess for thrombosis, including disseminated intravascular coagulation.
In the current issue of **Blood Research**, Han and Park report their retrospective experience with the use of SCBT to treat 5 bleeds in 4 hemophilia A patients with high-titer inhibitors and refractoriness to both bypassing agents used individually \[[@B10]\]. The bleeds were associated with various clinical conditions: peripherally inserted central catheter (PICC) insertion, small bowel surgery, hemothorax, and 2 total knee replacements. Patients were treated by infusing aPCC every 8 or 12 hours with administration of 1 or 2 doses of r-FVIIa between the doses of aPCC (50-100 IU/kg aPCC and 90 µg/kg r-FVIIa). The dosage and intervals were similar to those used in previous studies \[[@B7]\]. Twelve- to 24-hour bleeding episodes were controlled satisfactorily with no thrombotic events associated with SCBT. Because there is a paucity of data on the treatment of high-titer inhibitor patients, this report contributes valuable information for the management of bleeding episodes in such patients in Korea. It is hoped that a prospective, randomized, and controlled study of SCBT will be carried out in Korea in the near future to overcome the limits of this retrospective study.
| {
"pile_set_name": "PubMed Central"
} |
Text
====
Correction
----------
After publication of this article \[[@B1]\], we noticed that the acknowledgement of our funding source was not sufficiently specific. We hereby acknowledge that this research was supported in part by the Food and Drug Administration Office of Women\'s Health, USA.
| {
"pile_set_name": "PubMed Central"
} |
######
Comparison of the cost of antiretroviral treatment, by manufacturer and region
Manufacturer ARV (dose) No. of Sub-Saharan Africa countries (no. of individual transactions) Sub-Saharan Africa median cost per person per year (IQR), US\$ No. of Non-Africa countries (no. of individual transactions) Non-Africa median cost per person per year (IQR), US\$ Price rise, non-Africa vs. Sub-Saharan Africa (%)
-------------- ---------------------- ---------------------------------------------------------------------- ---------------------------------------------------------------- -------------------------------------------------------------- -------------------------------------------------------- ---------------------------------------------------
Branded ABC (300 mg) 2 (93) 315 (294--315) 3 (10) 547 (299--602) 74
ATV (300 mg) 2 (170) 357 (124--357) 2 (4) 1910 (1910--3496) 435
DRV (600 mg) 7 (84) 732 (732--806) 9 (31) 4690 (4075--5717) 541
LPV/r (200 mg/50 mg) 15 (492) 319 (272--374) 23 (128) 720 (456--932) 125
RAL (400 mg) 3 (52) 883 (883--1010) 1 (2) 3589 (3589--3589) 306
Generic ABC (300 mg) 18 (290) 192 (167--213) 33 (215) 178 (155--205) −7
ATV (300 mg) 6 (34) 296 (251--309) 14 (36) 245 (219--265) −17
DRV (600 mg) 2 (2) 990 (964--1016) 1 (1) 2964 (2964--2964) 199
LPV/r (200 mg/50 mg) 18 (164) 391 (282--429) 33 (187) 397 (349--435) 2
RAL (400 mg) 2 (28) 373 (373--634) 0 -- --
| {
"pile_set_name": "PubMed Central"
} |
Related literature {#sec1}
==================
For a related structure, see: Yamin *et al.* (2011[@bb6]). For standard bond lengths, see: Allen *et al.* (1987[@bb1]).
Experimental {#sec2}
============
{#sec2.1}
### Crystal data {#sec2.1.1}
C~11~H~12~ClFN~2~OS*M* *~r~* = 274.75Monoclinic,*a* = 14.818 (7) Å*b* = 10.291 (5) Å*c* = 18.201 (9) Åβ = 112.599 (12)°*V* = 2562 (2) Å^3^*Z* = 8Mo *K*α radiationμ = 0.46 mm^−1^*T* = 298 K0.50 × 0.22 × 0.07 mm
### Data collection {#sec2.1.2}
Bruker SMART APEX CCD area-detector diffractometerAbsorption correction: multi-scan (*SADABS*; Bruker, 2000[@bb2]) *T* ~min~ = 0.803, *T* ~max~ = 0.96913771 measured reflections4501 independent reflections2231 reflections with *I* \> 2σ(*I*)*R* ~int~ = 0.119
### Refinement {#sec2.1.3}
*R*\[*F* ^2^ \> 2σ(*F* ^2^)\] = 0.063*wR*(*F* ^2^) = 0.138*S* = 0.894501 reflections307 parametersH-atom parameters constrainedΔρ~max~ = 0.34 e Å^−3^Δρ~min~ = −0.30 e Å^−3^
{#d5e515}
Data collection: *SMART* (Bruker, 2000[@bb2]); cell refinement: *SAINT* (Bruker, 2000[@bb2]); data reduction: *SAINT*; program(s) used to solve structure: *SHELXTL* (Sheldrick, 2008[@bb4]); program(s) used to refine structure: *SHELXTL*; molecular graphics: *SHELXTL*; software used to prepare material for publication: *SHELXTL*, *PARST* (Nardelli, 1995[@bb3]) and *PLATON* (Spek, 2009[@bb5]).
Supplementary Material
======================
Crystal structure: contains datablock(s) global, I. DOI: [10.1107/S1600536811024743/rk2280sup1.cif](http://dx.doi.org/10.1107/S1600536811024743/rk2280sup1.cif)
Structure factors: contains datablock(s) I. DOI: [10.1107/S1600536811024743/rk2280Isup2.hkl](http://dx.doi.org/10.1107/S1600536811024743/rk2280Isup2.hkl)
Supplementary material file. DOI: [10.1107/S1600536811024743/rk2280Isup3.cml](http://dx.doi.org/10.1107/S1600536811024743/rk2280Isup3.cml)
Additional supplementary materials: [crystallographic information](http://scripts.iucr.org/cgi-bin/sendsupfiles?rk2280&file=rk2280sup0.html&mime=text/html); [3D view](http://scripts.iucr.org/cgi-bin/sendcif?rk2280sup1&Qmime=cif); [checkCIF report](http://scripts.iucr.org/cgi-bin/paper?rk2280&checkcif=yes)
Supplementary data and figures for this paper are available from the IUCr electronic archives (Reference: [RK2280](http://scripts.iucr.org/cgi-bin/sendsup?rk2280)).
The authors thank the Malaysian Government, Universiti Kebangsaan Malaysia, Universiti Malaysia Terengganu and the Ministry of Higher Education, Malaysia, for research grants UKM--GUP--NBT--08--27--110 and FRGS 59178.
Comment
=======
The title compound, is analogous to the previously reported *N*-(4-chlorobutanoyl)-*N*\'-(phenyl)thiourea (Yamin *et al.*, 2011) except the fluoro atom is attached at the *ortho*-position of the phenyl ring. The asymmetric unit consists of two independent molecules (Fig. 1). The whole molecule is not planar. However, the thiourea N1/C5/S1/N2/C6, N3/C16/S2/N4/C17 fragments and the benzene rings, (C6--C11) and (C17--C22) are each planar with maximum deviation of 0.020 (3)Å for N4 atom from the least square plane. In each molecule, the benzene ring and thiourea moiety forms dihedral angle of 74.78 (19)° and 82.3 (2)°, respectively. The same dihedral angle in *N*-(4-chlorobutanoyl)-N\'-(phenyl)thiourea are 72.98 (12)° and 81.47 (14)°, respectively. The bond lengths and angles are in normal ranges (Allen *et al.*, 1987) and comparable to those in the analog. Both molecules maintain their *trans*--*cis*-configuration with respect to the position of the butanoyl and fluorophenyl groups against the thiono C═S-group bond across the C---N bonds. Like in most of the carbonylthiourea derivatives, the classical intramolecular hydrogen bonds between the carbonyl oxygen atom and thioamide hydrogen atom, N2---H2···O1 and N4---H4···O2, in both molecules are present. In the crystal packing, the molecules are linked by N3---H3···S1^iii^ and N1---H1···S2^iii^; N2---H2···O2^i^ and N4---H4···O1^ii^; C14---H14A···S1^iii^ and C14---H14B···S2^iv^ intermolecular hydrogen bonds (symmetry codes as in Table 1) and form infinite chains along the *c*-axis (Fig. 2).
Experimental {#experimental}
============
A solution of 4-chlorobutanoylisothiocyanate (1.25 g, 6.33 mmol) in 30 ml acetone was added into a flask containing 30 ml acetone solution of 2-floroaniline (0.71 g, 6.33 mmol. The mixture was refluxed for 1 h. Then, the solution was filtered-off and left to evaporate at room temperature. The colourless solid was obtained after one day of evaporation (yield 83%, m.p. 411.7 K--415.5 K)
Refinement {#refinement}
==========
All H atoms attached to C and N atoms were fixed geometrically and treated as riding with C---H = 0.93Å or 0.97Å (aromatic and methylene) and N---H = 0.86Å (amino) with *U*~iso~(H) = 1.2*U*~eq~(C, N).
Figures
=======
{#Fap1}
{#Fap2}
Crystal data {#tablewrapcrystaldatalong}
============
------------------------- ---------------------------------------
C~11~H~12~ClFN~2~OS *F*(000) = 1136
*M~r~* = 274.75 *D*~x~ = 1.425 Mg m^−3^
Monoclinic, *P*2~1~/*c* Melting point = 411.7--415.5 K
Hall symbol: -P 2ybc Mo *K*α radiation, λ = 0.71073 Å
*a* = 14.818 (7) Å Cell parameters from 1404 reflections
*b* = 10.291 (5) Å θ = 1.5--25.0°
*c* = 18.201 (9) Å µ = 0.46 mm^−1^
β = 112.599 (12)° *T* = 298 K
*V* = 2562 (2) Å^3^ Plate, colourless
*Z* = 8 0.50 × 0.22 × 0.07 mm
------------------------- ---------------------------------------
Data collection {#tablewrapdatacollectionlong}
===============
------------------------------------------------------------ --------------------------------------
Bruker SMART APEX CCD area-detector diffractometer 4501 independent reflections
Radiation source: fine-focus sealed tube 2231 reflections with *I* \> 2σ(*I*)
graphite *R*~int~ = 0.119
Detector resolution: 83.66 pixels mm^-1^ θ~max~ = 25.0°, θ~min~ = 1.5°
ω scans *h* = −17→16
Absorption correction: multi-scan (*SADABS*; Bruker, 2000) *k* = −12→12
*T*~min~ = 0.803, *T*~max~ = 0.969 *l* = −13→21
13771 measured reflections
------------------------------------------------------------ --------------------------------------
Refinement {#tablewraprefinementdatalong}
==========
------------------------------------- -------------------------------------------------------------------------------------
Refinement on *F*^2^ Primary atom site location: structure-invariant direct methods
Least-squares matrix: full Secondary atom site location: difference Fourier map
*R*\[*F*^2^ \> 2σ(*F*^2^)\] = 0.063 Hydrogen site location: inferred from neighbouring sites
*wR*(*F*^2^) = 0.138 H-atom parameters constrained
*S* = 0.89 *w* = 1/\[σ^2^(*F*~o~^2^) + (0.0451*P*)^2^\] where *P* = (*F*~o~^2^ + 2*F*~c~^2^)/3
4501 reflections (Δ/σ)~max~ \< 0.001
307 parameters Δρ~max~ = 0.34 e Å^−3^
0 restraints Δρ~min~ = −0.30 e Å^−3^
------------------------------------- -------------------------------------------------------------------------------------
Special details {#specialdetails}
===============
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Geometry. All s.u.\'s (except the s.u. in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell s.u.\'s are taken into account individually in the estimation of s.u.\'s in distances, angles and torsion angles; correlations between s.u.\'s in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell s.u.\'s is used for estimating s.u.\'s involving l.s. planes.
Refinement. Refinement of *F*^2^ against ALL reflections. The weighted *R*-factor *wR* and goodness of fit *S* are based on *F*^2^, conventional *R*-factors *R* are based on *F*, with *F* set to zero for negative *F*^2^. The threshold expression of *F*^2^ \> σ(*F*^2^) is used only for calculating *R*-factors(gt) *etc*. and is not relevant to the choice of reflections for refinement. *R*-factors based on *F*^2^ are statistically about twice as large as those based on *F*, and *R*-factors based on ALL data will be even larger.
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å^2^) {#tablewrapcoords}
==================================================================================================
------ -------------- --------------- -------------- -------------------- --
*x* *y* *z* *U*~iso~\*/*U*~eq~
Cl1 0.49078 (11) 0.21242 (19) 0.00799 (10) 0.1269 (7)
F1 1.0669 (2) −0.1688 (3) 0.35144 (16) 0.0795 (9)
F2 0.3894 (2) 0.7354 (3) 0.12270 (17) 0.0868 (9)
S1 1.02276 (8) −0.03047 (12) 0.17159 (6) 0.0509 (3)
S2 0.24427 (8) 1.01331 (12) 0.02342 (6) 0.0502 (3)
Cl2 −0.23288 (9) 0.68710 (13) −0.21630 (7) 0.0804 (5)
O1 0.80228 (19) 0.1449 (3) 0.26210 (18) 0.0562 (9)
O2 0.0637 (2) 0.6917 (3) 0.07106 (17) 0.0552 (9)
N1 0.8543 (2) 0.0465 (3) 0.17333 (18) 0.0424 (9)
H1 0.8335 0.0212 0.1246 0.051\*
N2 0.9869 (2) 0.0693 (3) 0.29214 (19) 0.0428 (9)
H2 0.9467 0.1033 0.3103 0.051\*
N3 0.0918 (2) 0.8641 (3) 0.00426 (18) 0.0421 (9)
H3 0.0658 0.9086 −0.0388 0.051\*
N4 0.2217 (2) 0.8449 (3) 0.12431 (18) 0.0447 (9)
H4 0.1883 0.7860 0.1360 0.054\*
C1 0.5036 (3) 0.1269 (6) 0.0958 (3) 0.0858 (19)
H1A 0.4571 0.1608 0.1167 0.103\*
H1B 0.4885 0.0359 0.0831 0.103\*
C2 0.6047 (3) 0.1386 (5) 0.1580 (3) 0.0718 (16)
H2A 0.6195 0.2299 0.1700 0.086\*
H2B 0.6065 0.0966 0.2063 0.086\*
C3 0.6818 (3) 0.0806 (5) 0.1350 (2) 0.0547 (13)
H3A 0.6790 0.1208 0.0860 0.066\*
H3B 0.6682 −0.0113 0.1246 0.066\*
C4 0.7839 (3) 0.0961 (4) 0.1974 (3) 0.0453 (11)
C5 0.9539 (3) 0.0312 (4) 0.2162 (2) 0.0410 (11)
C6 1.0875 (3) 0.0553 (4) 0.3442 (2) 0.0407 (11)
C7 1.1249 (3) −0.0633 (5) 0.3737 (3) 0.0522 (12)
C8 1.2206 (4) −0.0794 (6) 0.4252 (3) 0.0739 (16)
H8 1.2448 −0.1613 0.4445 0.089\*
C9 1.2793 (4) 0.0285 (7) 0.4472 (3) 0.0822 (19)
H9 1.3443 0.0198 0.4816 0.099\*
C10 1.2431 (3) 0.1484 (6) 0.4192 (3) 0.0786 (17)
H10 1.2835 0.2210 0.4346 0.094\*
C11 1.1459 (3) 0.1625 (5) 0.3675 (3) 0.0610 (14)
H11 1.1209 0.2444 0.3489 0.073\*
C12 −0.2143 (3) 0.6210 (5) −0.1206 (3) 0.0691 (15)
H12A −0.2532 0.6696 −0.0977 0.083\*
H12B −0.2369 0.5316 −0.1269 0.083\*
C13 −0.1093 (3) 0.6249 (4) −0.0646 (2) 0.0530 (13)
H13A −0.0713 0.5723 −0.0865 0.064\*
H13B −0.1036 0.5862 −0.0145 0.064\*
C14 −0.0659 (3) 0.7590 (4) −0.0487 (2) 0.0490 (12)
H14A −0.0639 0.7934 −0.0977 0.059\*
H14B −0.1086 0.8147 −0.0334 0.059\*
C15 0.0351 (3) 0.7652 (4) 0.0149 (2) 0.0418 (11)
C16 0.1846 (3) 0.9006 (4) 0.0537 (2) 0.0386 (10)
C17 0.3156 (3) 0.8794 (4) 0.1821 (2) 0.0453 (12)
C18 0.3983 (3) 0.8244 (5) 0.1795 (3) 0.0566 (13)
C19 0.4902 (3) 0.8541 (6) 0.2342 (3) 0.0781 (17)
H19 0.5457 0.8151 0.2319 0.094\*
C20 0.4970 (4) 0.9431 (7) 0.2921 (4) 0.091 (2)
H20 0.5584 0.9657 0.3291 0.109\*
C21 0.4160 (4) 0.9992 (5) 0.2966 (3) 0.0777 (17)
H21 0.4221 1.0588 0.3366 0.093\*
C22 0.3243 (3) 0.9664 (4) 0.2409 (3) 0.0571 (13)
H22 0.2686 1.0039 0.2438 0.069\*
------ -------------- --------------- -------------- -------------------- --
Atomic displacement parameters (Å^2^) {#tablewrapadps}
=====================================
----- ------------- ------------- ------------- -------------- ------------- --------------
*U*^11^ *U*^22^ *U*^33^ *U*^12^ *U*^13^ *U*^23^
Cl1 0.0706 (11) 0.1612 (18) 0.1063 (14) 0.0426 (11) −0.0131 (9) 0.0101 (12)
F1 0.077 (2) 0.0548 (19) 0.098 (2) 0.0034 (15) 0.0232 (16) 0.0093 (16)
F2 0.068 (2) 0.111 (3) 0.081 (2) 0.0213 (17) 0.0283 (16) −0.0051 (19)
S1 0.0373 (6) 0.0695 (9) 0.0458 (7) 0.0083 (6) 0.0157 (5) −0.0007 (6)
S2 0.0383 (6) 0.0665 (9) 0.0432 (7) −0.0099 (6) 0.0128 (5) 0.0030 (6)
Cl2 0.0695 (9) 0.0838 (10) 0.0632 (9) −0.0010 (7) −0.0020 (7) −0.0005 (7)
O1 0.0393 (18) 0.072 (2) 0.051 (2) 0.0063 (15) 0.0105 (15) −0.0195 (17)
O2 0.0491 (19) 0.050 (2) 0.053 (2) −0.0104 (15) 0.0043 (15) 0.0103 (16)
N1 0.0292 (19) 0.059 (2) 0.035 (2) 0.0022 (17) 0.0072 (16) −0.0047 (16)
N2 0.0299 (19) 0.052 (2) 0.041 (2) 0.0042 (16) 0.0075 (16) −0.0039 (17)
N3 0.032 (2) 0.052 (2) 0.036 (2) −0.0023 (17) 0.0056 (16) 0.0003 (16)
N4 0.031 (2) 0.054 (2) 0.043 (2) −0.0076 (16) 0.0067 (16) 0.0077 (17)
C1 0.030 (3) 0.135 (5) 0.086 (4) −0.007 (3) 0.014 (3) −0.035 (4)
C2 0.037 (3) 0.110 (5) 0.065 (3) −0.002 (3) 0.017 (2) −0.014 (3)
C3 0.031 (2) 0.078 (4) 0.051 (3) 0.006 (2) 0.011 (2) −0.003 (2)
C4 0.037 (3) 0.051 (3) 0.045 (3) 0.003 (2) 0.013 (2) −0.001 (2)
C5 0.033 (2) 0.040 (3) 0.045 (3) −0.0013 (19) 0.009 (2) 0.003 (2)
C6 0.032 (2) 0.053 (3) 0.033 (2) 0.004 (2) 0.0073 (18) −0.004 (2)
C7 0.046 (3) 0.055 (3) 0.051 (3) 0.007 (3) 0.013 (2) −0.002 (2)
C8 0.058 (4) 0.089 (4) 0.061 (3) 0.032 (3) 0.007 (3) 0.008 (3)
C9 0.034 (3) 0.133 (6) 0.063 (4) 0.022 (3) 0.001 (3) −0.007 (4)
C10 0.036 (3) 0.100 (5) 0.089 (4) −0.013 (3) 0.011 (3) −0.026 (3)
C11 0.047 (3) 0.055 (3) 0.070 (3) 0.003 (2) 0.010 (2) −0.007 (3)
C12 0.050 (3) 0.082 (4) 0.068 (3) −0.022 (3) 0.015 (3) −0.010 (3)
C13 0.046 (3) 0.059 (3) 0.048 (3) −0.012 (2) 0.011 (2) −0.004 (2)
C14 0.032 (2) 0.052 (3) 0.060 (3) 0.003 (2) 0.014 (2) −0.004 (2)
C15 0.032 (2) 0.042 (3) 0.045 (3) 0.000 (2) 0.008 (2) −0.007 (2)
C16 0.028 (2) 0.048 (3) 0.038 (3) −0.0008 (19) 0.0109 (19) −0.003 (2)
C17 0.031 (3) 0.053 (3) 0.043 (3) −0.005 (2) 0.005 (2) 0.010 (2)
C18 0.045 (3) 0.069 (4) 0.053 (3) 0.000 (3) 0.016 (2) 0.005 (3)
C19 0.037 (3) 0.104 (5) 0.082 (4) 0.006 (3) 0.011 (3) 0.025 (4)
C20 0.047 (4) 0.105 (5) 0.089 (5) −0.022 (4) −0.007 (3) 0.022 (4)
C21 0.068 (4) 0.077 (4) 0.068 (4) −0.016 (3) 0.005 (3) −0.013 (3)
C22 0.050 (3) 0.053 (3) 0.058 (3) 0.001 (2) 0.009 (2) 0.000 (3)
----- ------------- ------------- ------------- -------------- ------------- --------------
Geometric parameters (Å, °) {#tablewrapgeomlong}
===========================
---------------- ----------- ------------------- -----------
Cl1---C1 1.769 (6) C6---C7 1.361 (6)
F1---C7 1.347 (5) C6---C11 1.365 (5)
F2---C18 1.350 (5) C7---C8 1.377 (6)
S1---C5 1.654 (4) C8---C9 1.372 (7)
S2---C16 1.675 (4) C8---H8 0.9300
Cl2---C12 1.791 (5) C9---C10 1.364 (7)
O1---C4 1.212 (4) C9---H9 0.9300
O2---C15 1.210 (4) C10---C11 1.393 (6)
N1---C4 1.374 (5) C10---H10 0.9300
N1---C5 1.388 (4) C11---H11 0.9300
N1---H1 0.8600 C12---C13 1.498 (5)
N2---C5 1.336 (5) C12---H12A 0.9700
N2---C6 1.433 (5) C12---H12B 0.9700
N2---H2 0.8600 C13---C14 1.502 (5)
N3---C16 1.375 (4) C13---H13A 0.9700
N3---C15 1.380 (5) C13---H13B 0.9700
N3---H3 0.8600 C14---C15 1.501 (5)
N4---C16 1.318 (4) C14---H14A 0.9700
N4---C17 1.430 (5) C14---H14B 0.9700
N4---H4 0.8600 C17---C22 1.363 (6)
C1---C2 1.496 (5) C17---C18 1.366 (6)
C1---H1A 0.9700 C18---C19 1.377 (6)
C1---H1B 0.9700 C19---C20 1.369 (7)
C2---C3 1.485 (6) C19---H19 0.9300
C2---H2A 0.9700 C20---C21 1.363 (8)
C2---H2B 0.9700 C20---H20 0.9300
C3---C4 1.511 (5) C21---C22 1.389 (6)
C3---H3A 0.9700 C21---H21 0.9300
C3---H3B 0.9700 C22---H22 0.9300
C4---N1---C5 129.5 (4) C9---C10---C11 120.2 (5)
C4---N1---H1 115.2 C9---C10---H10 119.9
C5---N1---H1 115.2 C11---C10---H10 119.9
C5---N2---C6 121.9 (4) C6---C11---C10 119.5 (5)
C5---N2---H2 119.0 C6---C11---H11 120.2
C6---N2---H2 119.0 C10---C11---H11 120.2
C16---N3---C15 128.2 (3) C13---C12---Cl2 112.4 (4)
C16---N3---H3 115.9 C13---C12---H12A 109.1
C15---N3---H3 115.9 Cl2---C12---H12A 109.1
C16---N4---C17 122.1 (3) C13---C12---H12B 109.1
C16---N4---H4 118.9 Cl2---C12---H12B 109.1
C17---N4---H4 118.9 H12A---C12---H12B 107.8
C2---C1---Cl1 111.9 (4) C12---C13---C14 114.3 (4)
C2---C1---H1A 109.2 C12---C13---H13A 108.7
Cl1---C1---H1A 109.2 C14---C13---H13A 108.7
C2---C1---H1B 109.2 C12---C13---H13B 108.7
Cl1---C1---H1B 109.2 C14---C13---H13B 108.7
H1A---C1---H1B 107.9 H13A---C13---H13B 107.6
C3---C2---C1 114.2 (4) C15---C14---C13 114.4 (3)
C3---C2---H2A 108.7 C15---C14---H14A 108.7
C1---C2---H2A 108.7 C13---C14---H14A 108.7
C3---C2---H2B 108.7 C15---C14---H14B 108.7
C1---C2---H2B 108.7 C13---C14---H14B 108.7
H2A---C2---H2B 107.6 H14A---C14---H14B 107.6
C2---C3---C4 113.7 (4) O2---C15---N3 123.0 (4)
C2---C3---H3A 108.8 O2---C15---C14 123.2 (4)
C4---C3---H3A 108.8 N3---C15---C14 113.8 (4)
C2---C3---H3B 108.8 N4---C16---N3 117.2 (4)
C4---C3---H3B 108.8 N4---C16---S2 123.2 (3)
H3A---C3---H3B 107.7 N3---C16---S2 119.7 (3)
O1---C4---N1 123.3 (4) C22---C17---C18 119.0 (4)
O1---C4---C3 124.0 (4) C22---C17---N4 120.8 (4)
N1---C4---C3 112.6 (4) C18---C17---N4 120.3 (4)
N2---C5---N1 115.8 (4) F2---C18---C17 118.9 (4)
N2---C5---S1 124.8 (3) F2---C18---C19 118.9 (5)
N1---C5---S1 119.4 (3) C17---C18---C19 122.2 (5)
C7---C6---C11 119.3 (4) C20---C19---C18 117.7 (5)
C7---C6---N2 120.9 (4) C20---C19---H19 121.2
C11---C6---N2 119.8 (4) C18---C19---H19 121.2
F1---C7---C6 119.2 (4) C21---C20---C19 121.6 (5)
F1---C7---C8 118.6 (5) C21---C20---H20 119.2
C6---C7---C8 122.2 (5) C19---C20---H20 119.2
C9---C8---C7 118.2 (5) C20---C21---C22 119.4 (5)
C9---C8---H8 120.9 C20---C21---H21 120.3
C7---C8---H8 120.9 C22---C21---H21 120.3
C10---C9---C8 120.6 (5) C17---C22---C21 120.2 (5)
C10---C9---H9 119.7 C17---C22---H22 119.9
C8---C9---H9 119.7 C21---C22---H22 119.9
---------------- ----------- ------------------- -----------
Hydrogen-bond geometry (Å, °) {#tablewraphbondslong}
=============================
---------------------- --------- --------- ----------- ---------------
*D*---H···*A* *D*---H H···*A* *D*···*A* *D*---H···*A*
N2---H2···O1 0.86 2.02 2.691 (5) 134
N2---H2···O2^i^ 0.86 2.40 3.128 (5) 143
N4---H4···O2 0.86 2.02 2.676 (5) 133
N4---H4···O1^ii^ 0.86 2.32 3.033 (5) 141
N3---H3···S1^iii^ 0.86 2.59 3.447 (4) 176
N1---H1···S2^iii^ 0.86 2.52 3.364 (4) 169
C3---H3A···Cl1 0.97 2.76 3.189 (5) 107
C14---H14A···Cl2 0.97 2.82 3.190 (4) 103
C14---H14A···S1^iii^ 0.97 2.96 3.784 (5) 143
C14---H14B···S2^iv^ 0.97 2.74 3.691 (5) 168
---------------------- --------- --------- ----------- ---------------
Symmetry codes: (i) −*x*+1, *y*−1/2, −*z*+1/2; (ii) −*x*+1, *y*+1/2, −*z*+1/2; (iii) −*x*+1, −*y*+1, −*z*; (iv) −*x*, −*y*+2, −*z*.
###### Hydrogen-bond geometry (Å, °)
*D*---H⋯*A* *D*---H H⋯*A* *D*⋯*A* *D*---H⋯*A*
---------------------- --------- ------- ----------- -------------
N2---H2⋯O1 0.86 2.02 2.691 (5) 134
N2---H2⋯O2^i^ 0.86 2.40 3.128 (5) 143
N4---H4⋯O2 0.86 2.02 2.676 (5) 133
N4---H4⋯O1^ii^ 0.86 2.32 3.033 (5) 141
N3---H3⋯S1^iii^ 0.86 2.59 3.447 (4) 176
N1---H1⋯S2^iii^ 0.86 2.52 3.364 (4) 169
C14---H14*A*⋯S1^iii^ 0.97 2.96 3.784 (5) 143
C14---H14*B*⋯S2^iv^ 0.97 2.74 3.691 (5) 168
Symmetry codes: (i) ; (ii) ; (iii) ; (iv) .
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1-1}
============
Previous reports have clearly documented that aortic stiffness is associated not only with coronary artery disease (CAD) but also with cardiovascular risk factors, such as hypercholesterolemia, smoking, older age, hypertension, impaired glucose tolerance and diabetes mellitus (DM), and obesity ([@ref1]-[@ref8]). Increased aortic stiffness has been reported as an independent predictor of cardiovascular mortality and morbidity ([@ref9], [@ref10]) as well as of CAD severity ([@ref11],[@ref12]). Atherosclerosis-related cardiovascular diseases have a clinical course starting from endothelial dysfunction and progressing to clinically evident diseases. In addition to the prediction of cardiovascular events in a patient with established cardiovascular disease, rationale for evaluating aortic stiffness includes early detection and risk stratification of a subject at a risk of cardiovascular disease. Therefore, current guidelines recommend the assessment of aortic stiffness by the aortic pulse-wave velocity (PWV) method to evaluate cardiovascular risk ([@ref13]). Although PWV is considered as the gold-standard method to determine arterial stiffness, the necessities of trained medical staff and special devices reduce its potential usage. Accordingly, other echocardiography-derived stiffness parameters, such as aortic strain (AS), aortic distensibility (AD), stiffness index, augmentation index, and more recently, color M-mode derived propagation velocity of the descending thoracic aorta (APV), have been analyzed in the context of the association between aortic stiffness and CAD in numerous studies ([@ref11],[@ref12], [@ref14]-[@ref17]).
AD and AS, which have been widely used as noninvasive aortic stiffness parameters in clinical trials, reflect the mechanical properties and elasticity of the aorta. AS reflects the percent change in vessel diameter, whereas AD denotes the effect of pulse pressure in any alteration in the vessel diameter. AS and AD measurements do not require special devices and trained staff; however, they require some additional computations.
Evolved recently, APV is considered as a practical and sensitive aortic stiffness parameter ([@ref17]). The clinical value of APV in patients with coronary and carotid atherosclerosis and other associated clinical conditions, such as endothelial dysfunction and hypertension, has been the subject of several studies ([@ref17]-[@ref22]). In addition, the association of APV with other stiffness parameters has been studied in conditions related to increased stiffness, such as obstructive sleep apnea syndrome and newly diagnosed type 2 DM ([@ref23], [@ref24]). However, no study has evaluated these parameters in low-risk groups. Accordingly, the present study aimed to evaluate the correlation of APV with the aortic stiffness parameters AS and AD in a population with a relatively low risk of cardiovascular disease. This cohort can also be considered as candidates to benefit from early detection and risk stratification for cardiovascular disease. We also evaluated the association of each studied parameters, such as age, obesity, and blood pressure, with known risk factors for increased stiffness.
Methods {#sec1-2}
=======
Study population {#sec2-1}
----------------
This cross-sectional study consisted of 97 consecutive healthy individuals (50 males and 47 females) undergoing routine health check-up at the Department of Cardiology Outpatient Clinic at Süleyman Demirel University School of Medicine between July and December in 2015. Clinical data obtained from patient interviews included past medical history and classical cardiovascular risk factors. Exclusion criteria were as follows: age of ≥50 years, known hypertension and DM, left ventricular (LV) ejection fraction of \<50%, severe diastolic dysfunction (E/Em \>15), moderate-to-severe valvular heart disease, arrhythmia, aortic disease, known coronary artery disease or presence of two or more classical risk factors for CAD, chronic inflammatory or connective tissue disease, chronic pulmonary disease, renal failure (GFR \<90 ml/min/1.73 m^2^) and inadequate acoustic windows. These parameters were considered as exclusion criteria because these conditions might have influenced the outcomes of some of the measured parameters. None of the participants were taking any medications.
Hypertension, DM, and hypercholesterolemia were either self-reported or determined on the basis of current treatment. Diagnostic ranges for newly diagnosed hypertension (\>140/90 mm Hg), DM (fasting glucose level \>126 mg/dL), and hyperlipidemia (total cholesterol \>200 mg/dL or triglycerides \>150 mg/dL) were formed according to existing guidelines. Severe diastolic dysfunction was determined using the echocardiographic examination mainly on the basis of tissue Doppler imaging (TDI) method (E/Em \>15). Positive family history of CAD was defined as CAD diagnosed in a first-degree relative before the age of 55 years for men and 65 years for women. Individuals who had smoked more than 100 cigarettes in their entire lifetime were considered as smokers. Echocardiographic examinations were performed at least 2 h after the smokers in the cohort smoked the last cigarette. Waist circumference was measured at the level of the umbilicus in the standing position. All participants provided informed consent to participate in this study. The Institutional Ethics Committee approved this study.
Transthoracic echocardiographic examination {#sec2-2}
-------------------------------------------
All subjects underwent detailed echocardiographic examination by a single experienced cardiologist using a Philips I-E33 imaging system (Andover, Massachusetts, USA) with a 2.5 MHz transducer. Conventional echocardiographic measurements were obtained using two-dimensional, color flow, pulse- and continuous-wave Doppler modalities. The longitudinal mitral annular velocities from the apical four-chamber view were recorded using TDI echocardiography for further assessment of LV diastolic function. Systolic and diastolic diameters of the ascending aorta and aortic propagation velocity were recorded. All echocardiographic measurements were obtained for at least three consecutive beats and averaged.
Aortic diameter measurements were obtained in the parasternal long-axis view by placing the M-mode sampling line 3 cm above the aortic valve ([@ref18]). Diastolic diameter (DD) was measured at the peak of the QRS complex of the simultaneously recorded electrocardiogram, and systolic diameter (SD) was measured at the time of full opening of the aortic valve ([Fig. 1a](#F1){ref-type="fig"}).
{#F1}
AS and AD were calculated according to previously defined formulas using aortic diameters and blood pressure as follows ([@ref18]):
Concomitantly, blood pressure was measured using the ausculatory method after the echocardiographic study, ensuring that the subjects had rested for at least 5 min.
APV measurement was performed using color M-mode Doppler recordings obtained from the suprasternal window in a supine position, with the cursor parallel to the main flow of direction in the descending aorta, as previously described ([@ref17]). The color Doppler Nyquist limit was adapted to 30--50 cm/s, with switching to the M-mode with a recorder sweep rate of 150 mm/s. Subsequently, aliasing velocity was shifted until a clear delineation of the flame-shaped isovelocity map slope was observed. APV was calculated by dividing the distance between the beginning and end of the slope to the corresponding time ([Fig. 1b](#F1){ref-type="fig"}).
Statistics {#sec2-3}
----------
Statistical analysis was performed using IBM SPSS Statistics for Windows, version 23 (IBM Corp., Armonk, N.Y., USA). Descriptive statistics were presented as mean ± standard deviation for continuous variables and as percentage for categorical data. Spearman correlation analysis was performed to assess the correlations of stiffness parameters with the other study variables. A two-tailed p value of \<0.05 was considered as statistically significant.
Results {#sec1-3}
=======
The study population (51.5% male; mean age, 35.1±8.1 years) had a relatively low-cardiovascular risk profile with the total cholesterol level of 177±39.1 mg/dL, fasting glucose level of 97±14.6 mg/dL, and systolic blood pressure (SBP) and diastolic blood pressure (DBP) of 108.6±15.1 and 69.1±11.3 mm Hg, respectively. None of the participants were hypertensive, and four participants with newly diagnosed DM were excluded from the study. The patients had slightly high BMI (26.9±4.5 kg/m^2^) and waist circumference of 90.6±12.6 cm, along with a high rate of smoking (34%). Demographic, clinical, laboratory, and echocardiographic findings of the study population are presented in [Table 1](#T1){ref-type="table"}. Age was negatively correlated with AS and AD, and the relationship was statistically significant (r=--0.31, p\<0.01 for AS and r=--0.29, p=0.01 for AD). Significant correlations were also present between SBP and AD (r=--0.36, p\<0.01) and DBP and AS (p=--0.27, p\<0.01). Waist circumference and BMI were negatively correlated with AS (r=--0.31, p\<0.01 and r=--0.21, p=0.04, respectively) and with AD (r=--0.37, p\<0.01 and r=--0.25--4, p=0.02, respectively); these relationships were also statistically significant. However, no significant correlation of APV with age, SBP, DBP, waist circumference, and BMI was identified (r=0.1, --0.04, --0.04, --0.08, 0.08, respectively, p\>0.05 for all comparisons) ([Table 2](#T2){ref-type="table"}). No significant correlation of APV with AS and AD could be identified (r=--0.05, p=0.6 and r=--0.17, p=0.8, respectively).
######
Demographic, clinical and echocardiographic findings of the study population (n=97)
Mean±SD Range n (%)
----------------------------------- ------------------------- ------- ------------
Age, year 35.1±8.1 20-50
Sex, female 474 (8.5%)
BMI, kg/m^2^ 26.9±4.5
Waist circumference, cm 90.6±12.6
SBP, mm Hg 108.6±15.1
DBP, mm Hg 69.1±11.3
Medical history at admission
Hypertension 0 (0%)
Diabetes mellitus 0 (0%)
Hyper cholesterolemia 24 (24.7%)
Hyper triglyceridemia 13 (13.5%)
Smoking 33 (34%)
Positive family history for IHD 4 (4%)
Kreatinin, mg/dL 0.93±0.14
Glukoz, mg/dL 97±14.6
Total cholesterol, mg/dL 177±39.1
LDL cholesterol, mg/dL 105.1±34.6
HDL cholesterol, mg/dL 48.9±24
Haemoglobin, gr/dL 14.4±1.6
Triglicerid 136±90
WBC, 10^3^/mm^3^ 7.48±1.83
Aortic strain, % 12.87±6.39\*
11.11 (8.38--18.09)\*\*
AD, cm^2^ dyn^-1^ 10^-3^ 0.68±0.39\*
0.59 (0.39--0.87)\*\*
APV, cm/s 62.9±29.5\*
57.3 (40.2--76)\*\*
LVEF, % 64±5.5
LA, mm 32.1±4.3
Mitral E/Em 5.6±1.75
\#Diastolic dysfunction, E/Em \>8 10 (10%)
Data is presented as mean ± standard deviation\* and median (interquartile range, 25th-75th)\*\*. \# Diastolic dysfunction was determined using the echocardiographic examination based on the tissue Doppler imaging method (E/Em \> 8).
AD: aortic distensibility; APV: aortic propagation velocity; BMI: body mass index; DBP: diastolic blood pressure; Em: early diastolic mitral annular velocity; HDL: high-density lipoprotein cholesterol; LA: left atrium; LDL: low-density lipoprotein cholesterol; LVEF: left ventricular ejection fraction; Mitral E: early mitral inflow velocity; SBP: systolic blood pressure; WBC: white blood cell
######
The correlations of stiffness parameter with in themselves and with age, blood pressure, BMI, waist circumference
Aortic strain Aortic distensibility APV
----- --------------- ------------------------------------------- --------- ------------------------------------------- --------- -------
APV r=-0.05 p=0.6 r=-0.17 p=0.8
Age r=-0.31 p\<0.01[\*\*](#t2f2){ref-type="table-fn"} r=-0.29 p=0.01[\*](#t2f1){ref-type="table-fn"} r=0.1 p=0.3
SBP r=-0.14 p=0.2 r=-0.36 p\<0.01[\*\*](#t2f2){ref-type="table-fn"} r=-0.04 p=0.7
DBP r=-0.27 p\<0.01[\*\*](#t2f2){ref-type="table-fn"} r=-0.16 p=0.1 r=-0.04 p=0.7
BMI r=-0.21 p=0.04[\*](#t2f1){ref-type="table-fn"} r=-0.24 p=0.02[\*](#t2f1){ref-type="table-fn"} r=-0.08 p=0.4
WC r=-0.31 p\<0.01[\*\*](#t2f2){ref-type="table-fn"} r=-0.37 p\<0.01[\*\*](#t2f2){ref-type="table-fn"} r=0.08 p=0.4
Spearman correlation analysis was used for statistical analyses.
Correlation is significant at the 0.05 level (2-tailed);
Correlation is significant at the 0.01 level (2-tailed).
AVP-color M-Mode propagation velocity of descending aorta; BMI-body mass index, DBP-diastolic blood pressure, SBP-systolic blood pressure, WC-Waist circumference
The reproducibility of aortic diameter measurements during systole and diastole was higher than that of APV, as shown by the intra-class correlation coefficient test for intra-observer variability (intra-class correlation coefficient = 0.96, p\<0.001; intra-class correlation coefficient =0.97, p\<0.001; and intra-class correlation coefficient = 0.90, p=0.001, respectively). Additional tests in 15 study participants showed lower but acceptable inter-observer variability for APV (r=0.50, p=0.02).
Discussion {#sec1-4}
==========
Although previous studies have clearly demonstrated a relationship of classic cardiovascular risk factors with AS and AD in high-risk populations, there is no substantial data on this association in low-risk group. The present observational study demonstrated an association of classical cardiovascular risk factors with AS and AD in a low-risk study group. However, an association of these risk factors with APV was not observed. In addition, no significant correlation of APV with AS or AD was detected in this low-cardiovascular risk population.
Aortic stiffness reflecting arterial stiffness describes the rigidity of the arteries and is mainly determined by the arterial wall components, including extracellular matrix, endothelial cells, smooth muscle cells, and other functional elements of the vessel wall ([@ref25]). Pathophysiological alterations affecting these components in the context of cardiovascular risk factors induce structural alterations leading to increased arterial stiffness before the development of CVD ([@ref1]-[@ref12]). Therefore, the evaluation of aortic stiffness is a useful for detecting CVD earlier than the occurrence of clinically evident disease in risky groups as well as for predicting adverse clinical outcomes in patients with established disease ([@ref9], [@ref10], [@ref14], [@ref26], [@ref27]). Measurement of aortic stiffness beyond established risk stratification strategies with classical risk factors, such as age, gender, hypertension, hyperlipidemia, smoking, and obesity, may lead to more appropriate allocation of patient to the correct risk group, such as the intermediate-to-high-risk group, instead of erroneous allocation to a low-to-intermediate-risk group ([@ref9], [@ref13], [@ref28]). Therefore, the determination of aortic stiffness is of greater importance in low- or intermediate-risk populations.
Although several ultrasonographic indices have been studied for quantifying aortic stiffness, carotid--femoral PWV is the most validated modality for the noninvasive evaluation of aortic stiffness and is highly recommended for clinical implementation ([@ref9], [@ref13]). However, the requirement of trained medical staff and special devices reduces its clinical applicability. For diagnostic methods to be widely applicable, they should have high predictive value, availability, reproducibility, and cost-effectiveness. Therefore, it is important to know the performance of other aortic stiffness indices in the preclinical patient population before the establishment of frank disease. AD and AS are echocardiography- and magnetic resonance imaging (MRI)-derived elasticity indices of the aorta and have been shown to be well correlated with PWV measurements ([@ref29]-[@ref31]).
An association between age and aortic stiffness is well documented ([@ref9], [@ref11], [@ref32]). A reduction in aortic elasticity with increasing age has been reported even in the absence of overt cardiovascular disease ([@ref14], [@ref29]). In the present study, we demonstrated an inverse correlation between the aortic elasticity indices, AS and AD, and age in a low-cardiovascular risk population, which is consistent with the findings in previous reports. However, we did not detect a correlation between APV and age ([Fig. 2](#F2){ref-type="fig"}). A previous study has suggested that individuals younger than 50 years with no overt cardiovascular disease have a more explicit association between MRI-derived AD and age than between PWV and age ([@ref29]). The increases in MRI-derived cf-PWV and aortic arch PWV were more prominent in patients older than 50 years than in younger patients in the same study.
{#F2}
Obesity is another well-known risk factor for increased vascular stiffness and CVD ([@ref33]). Increased arterial stiffness has been reported even in obese children ([@ref29], [@ref34], [@ref35]), and weight loss has been shown to improve arterial compliance ([@ref33], [@ref36]). Our results are consistent with those in these previous reports with respect to indicating significant associations of AS and AD with BMI as well as of AS and AD with waist circumference reflecting visceral adiposity. However, no association of APV with BMI or waist circumference was observed.
There is a bidirectional association between aortic stiffness and HT ([@ref37]). Accordingly, the relationship between aortic stiffness and HT can implicate the rational association between HT and the other cardiovascular risk factors, such as smoking, hypercholesterolemia, and DM. In the present study, inverse correlations were observed between AS and DBP as well as between AD and SBP even in individuals with normal blood pressure. However, no correlation was observed between APV and blood pressure. A significant correlation between blood pressure and APV has been reported in two studies on hypertensive patients ([@ref21], [@ref22]). In these studies, higher APV values were noted in the normotensive control group than in the hypertensive patient group. However, these studies did not report any correlation analysis between APV and blood pressure in the normotensive control group.
In a recent study, APVs were lower in patients with CAD than in those without CAD and were significantly correlated with the echocardiography-derived aortic stiffness parameters (r=0.556, p\<0.001 for AS and r=0.483, p\<0.001 for AD). However, this study did not report any similar analysis in the control group ([@ref18]). In another study, APV was found to be inversely correlated with PWV (r=--0.580, p\<0.001), and positively correlated with brachial artery flow-mediated dilatation (r=0.564, p\<0.001) in patients with obstructive sleep apnea syndrome having a relatively higher cardiovascular risk profile ([@ref23]). In our study, we observed no statistically significant correlation between APV and the conventional aortic stiffness parameters AS and AD (r=--0.05, p=0.6 and r=--0.17, p=0.8, respectively) in healthy individuals ([Fig. 3](#F3){ref-type="fig"}). The lack of correlation between these parameters might be explained by different mechanisms of blood pressure propagation, which drive aortic elasticity indices and flow propagation. Pathologic conditions, such as CAD, may have a composite effect on these entities.
{#F3}
Study limitations {#sec2-4}
-----------------
The major limitation of the present study is that the association of APV was not evaluated using a more validated aortic stiffness evaluation method, such as PWV, instead of or in addition to the evaluation of AS and AD. However, the main aim of our study was to evaluate this association using more practical, easily applicable, and safe methods. On the other hand, APV is a newly established stiffness parameter that has been extensively studied in established CAD populations. Although the cohort in the present study had a low-risk profile for atherosclerosis, we did not investigate the presence of atherosclerosis using validated invasive angiographic or noninvasive imaging methods, such as carotid intima-media thickness measurement or multidetector row computed tomography. The small size and low-risk profile of the study population are other possible limitations that might also have affected the statistical analysis.
Conclusion {#sec1-5}
==========
APV has been suggested as a novel and simple echocardiographic parameter of aortic stiffness in relatively high-risk and established CVD groups. The association of APV with other stiffness parameters in low-risk groups has not yet been evaluated. The present study showed lack of association between APV and AS, as well as between APV and AD. No association was observed between APV and known classical risk factors related to increased stiffness, such as age, BMI, waist circumference, and blood pressure, in the low-cardiovascular risk group. On the other hand, the associations of AS and AD with age, BMI, waist circumference, and blood pressure were statistically significant. The validity of APV should be further investigated considering its lower reproducibility compared with that of the classical stiffness parameters.
The author(s) declared no financial support and potential conflict of interest with respect to the research, authorship, and/or publication of this article.
**Conflict of interest:** None declared.
**Peer-review:** Externally peer-reviewed.
**Authorship contributions:** Concept -- H.A., D.E., Y.T.; Design -- H.A., Y.T.; Supervision -- F.K., Y.T., D.E.; Materials -- F.K., S.G., H.A.; Data collection &/or processing -- H.A., S.G., F.K., H.A.; Analysis &/or interpretation -- H.A.B., Y.T.; Literature search -- H.A., D.E.; Writing -- H.A., F.K., S.G.; Critical review -- D.E., Y.T.
| {
"pile_set_name": "PubMed Central"
} |